agent build

skills/activity-analysis.md

@@ -0,0 +1,99 @@
+---
+id: activity-analysis
+name: Activity Analysis
+description: Break down what has happened in this workspace, by kind of event and by account.
+category: operations
+pages:
+  - activity
+status: active
+version: 1
+triggers:
+  - activity breakdown
+  - event breakdown
+  - what kind of events
+  - events by type
+  - who did what
+  - busiest account
+owliver:
+  enabled: true
+  suggestions:
+    - label: What kinds of event are there?
+      capability: table
+    - label: Summarize workspace activity
+      capability: summary
+    - label: Activity over recent periods
+      capability: flow
+  capabilities:
+    - summary
+    - stats
+    - table
+    - list
+    - progress
+    - flow
+  responses:
+    summary:
+      title: Activity breakdown
+      source: activity.breakdown
+    stats:
+      title: Activity breakdown
+      source: activity.breakdown
+    table:
+      title: Events by kind
+      source: activity.breakdown
+    list:
+      title: Events by kind
+      source: activity.breakdown
+    progress:
+      title: Events by kind
+      source: activity.breakdown
+    flow:
+      title: Activity over time
+      source: activity.breakdown
+      periods:
+        - last-7-days
+        - this-month
+        - previous-month
+---
+
+# Activity Analysis
+
+## Purpose
+
+- Report what has happened in this workspace and in what proportion.
+- Count how many accounts are active.
+- Show activity across recent periods.
+
+## Capabilities
+
+- Break events down by kind, with each kind's share.
+- Count distinct event kinds and active accounts.
+- Window the breakdown by period.
+
+## Data
+
+Reads `activity.breakdown`, which counts `UserActivity` records by `event_type`
+and by account.
+
+## Analysis
+
+Events are counted by kind and expressed as a share of the total, because a raw
+count means little without knowing whether twelve logins is most of the log or a
+fraction of it.
+
+Stored event names are machine keys; they are rendered as words so a reader does
+not have to translate `hire_candidate` in their head.
+
+## Output
+
+Total events, number of distinct kinds, number of active accounts, then a row
+per kind with its count and share.
+
+## Limitations
+
+- This describes the audit log, not the underlying records. Ten `apply_job`
+  events mean ten logged actions, which is not a guarantee of ten applications
+  surviving in the pipeline.
+- Overlapping periods are deduplicated by event, so asking for today and the
+  last seven days together does not double-count today.
+- This counts activity; it does not judge it. Whether a pattern is unusual is
+  Anomaly Detection's question.

skills/analytics-insights.md

@@ -0,0 +1,27 @@
+---
+id: analytics-insights
+name: Analytics Insights
+description: Help Owliver explain the analytics shown on the current page.
+pages:
+  - analytics
+status: active
+actions:
+  - navigate_to_analytics
+---
+
+# Analytics Insights
+
+## Purpose
+
+Explain the figures on the Analytics page — conversion, speed, department
+performance — using the same records the page renders.
+
+## Capabilities
+
+- Explain hiring trend and conversion.
+- Compare department performance.
+- Identify where the funnel loses candidates.
+
+## Actions
+
+- navigate_to_analytics

skills/anomaly-detection.md

@@ -0,0 +1,96 @@
+---
+id: anomaly-detection
+name: Anomaly Detection
+description: Surface activity that departs from this workspace's own pattern — and stay quiet when nothing does.
+category: operations
+pages:
+  - activity
+  - control-center
+status: active
+version: 1
+triggers:
+  - anomaly
+  - anomalies
+  - anomalous
+  - unusual
+  - out of pattern
+  - suspicious
+owliver:
+  enabled: true
+  suggestions:
+    - label: Is anything unusual?
+      capability: insight
+    - label: Show the signals
+      capability: table
+  capabilities:
+    - summary
+    - insight
+    - list
+    - table
+    - stats
+  responses:
+    summary:
+      title: Activity signals
+      source: activity.signals
+    insight:
+      title: Unusual activity
+      source: activity.signals
+    list:
+      title: Signals
+      source: activity.signals
+    table:
+      title: Signals
+      source: activity.signals
+    stats:
+      title: Activity signals
+      source: activity.signals
+---
+
+# Anomaly Detection
+
+## Purpose
+
+- Surface activity that departs from this workspace's own baseline.
+- Explain each signal rather than only naming it.
+- Report nothing when nothing departs, so a signal keeps its meaning.
+
+## Capabilities
+
+- Detect concentration, bursts, off-hours activity, silence and privileged-action share.
+- Report how many signals are currently raised.
+- Explain what each one means.
+
+## Data
+
+Reads `activity.signals`, which is the same detection the assistant's own
+greeting counts — one implementation in `lib/activitySignals.js`, so "two
+unusual patterns" means the same two wherever it is said.
+
+## Analysis
+
+Five patterns are checked against this workspace's own history:
+
+1. **Concentration** — one account is responsible for half or more of events.
+2. **Burst** — more than three actions from one account inside one hour.
+3. **Off-hours** — activity before 06:00 or after 22:00.
+4. **Silent** — a log that has events but nothing in the last 24 hours.
+5. **Privileged share** — more than 30% of events change who is employed or
+   what is being hired for.
+
+Only patterns that clear their threshold are reported. A workspace with nothing
+unusual returns no signals, not a low-severity note.
+
+## Output
+
+A count of raised signals, and one row per signal explaining what triggered it
+with the figure behind it.
+
+## Limitations
+
+- **A signal is a deviation from a baseline, not a verdict.** On a live
+  deployment most resolve to an integration, a bulk import or a busy afternoon.
+  Nothing here asserts wrongdoing.
+- Thresholds are fixed, not learned. A workspace whose normal pattern is one
+  busy account will report concentration every time it is asked.
+- The baseline is the whole activity log, not a rolling window, so a young
+  workspace has little to compare against.

skills/attendance-analysis.md

@@ -0,0 +1,105 @@
+---
+id: attendance-analysis
+name: Attendance Analysis
+description: Analyse workforce attendance, lateness and absence, and compare people and departments.
+category: workforce
+pages:
+  - analytics
+  - control-center
+status: active
+version: 1
+triggers:
+  - attendance
+  - absence
+  - absences
+  - absenteeism
+  - late arrival
+  - missed shift
+  - missed shifts
+owliver:
+  enabled: true
+  suggestions:
+    - label: How is attendance?
+      capability: summary
+    - label: Compare attendance by person
+      capability: table
+    - label: Attendance over recent periods
+      capability: flow
+  capabilities:
+    - summary
+    - stats
+    - table
+    - progress
+    - flow
+    - insight
+  responses:
+    summary:
+      title: Attendance
+      source: workforce.attendance
+    stats:
+      title: Attendance
+      source: workforce.attendance
+    table:
+      title: Attendance by person
+      source: workforce.attendance
+    progress:
+      title: Attendance by person
+      source: workforce.attendance
+    flow:
+      title: Attendance over time
+      source: workforce.attendance
+      periods:
+        - last-7-days
+        - this-month
+        - previous-month
+    insight:
+      title: Attendance
+      source: workforce.attendance
+---
+
+# Attendance Analysis
+
+## Purpose
+
+- Report how reliably the workforce is turning up.
+- Separate turning up from turning up on time, because they have different causes.
+- Compare people and departments so a problem can be located rather than only counted.
+
+## Capabilities
+
+- Summarize attendance, punctuality and missed shifts.
+- Compare attendance per person, worst first.
+- Show attendance across recent periods.
+
+## Data
+
+Reads `workforce.attendance`, which counts `ShiftRecord` entries — every shift
+scheduled, whether it was worked, how late it started and how long it ran.
+Department is the position's `role_category`, the same field Hired History and
+Analytics group by.
+
+## Analysis
+
+Attendance is the share of scheduled shifts that were **turned up for at all**,
+late or not. Punctuality is reported separately, as the share turned up for on
+time. Folding the two together would make a reliably-late team look absent and
+a genuinely absent one look better than it is.
+
+Minutes lost to lateness are reported alongside the count, because four late
+arrivals says nothing about whether it cost ten minutes or two hours.
+
+## Output
+
+Headline attendance and punctuality rates, missed shifts split into absences and
+no-shows, and a row per person with their record. Over periods, one figure per
+window.
+
+## Limitations
+
+- Only shifts that were scheduled are counted. Unrostered work does not appear.
+- A no-show and an absence are counted separately but both reduce attendance;
+  the distinction is in the detail line, not in the headline rate.
+- The roster is whoever has shift records. This workspace has three, so a
+  department average is one person's record — the per-person view is the more
+  honest read at this size.
+- Excused absence is a recognised status but none is currently recorded.

skills/bartending-training.md

@@ -0,0 +1,37 @@
+---
+id: bartending-training
+name: Bartending Training
+description: Training path for specs, speed and responsible service behind a bar.
+skill: bartending
+pages:
+  - positions
+  - profile
+---
+
+# Bartending Training
+
+Speed, specs, and the judgement to run a bar alone.
+
+Attached to Positions and Profile but not Candidates: bar levels are read when
+matching someone to a role and when they are planning their own development, and
+the recruiter's candidate view is kept to the skills the roles on file ask for.
+
+## Beginner
+
+Pour to spec and keep a station clean through a service.
+
+## Intermediate
+
+Refuse service without a scene, and document what happened.
+
+## Advanced
+
+Run a bar alone through a full event.
+
+## Expert
+
+Design a list and train the people who pour it.
+
+## Verification
+
+Owliver evaluates the recorded pour and the spoken refusal.

skills/candidate-analysis.md

@@ -0,0 +1,98 @@
+---
+id: candidate-analysis
+name: Candidate Analysis
+description: Analyse the applicant pool's quality, and how much of it has actually been screened.
+category: hiring
+pages:
+  - candidates
+  - candidates-analysis
+status: active
+version: 1
+triggers:
+  - candidate quality
+  - quality of candidates
+  - score band
+  - score bands
+  - screening coverage
+  - how strong*candidates
+  - how good*candidates
+owliver:
+  enabled: true
+  suggestions:
+    - label: How strong is the candidate pool?
+      capability: summary
+    - label: Show candidates by score
+      capability: table
+    - label: Score bands
+      capability: progress
+  capabilities:
+    - summary
+    - stats
+    - table
+    - list
+    - progress
+    - insight
+  responses:
+    summary:
+      title: Candidate quality
+      source: candidates.quality
+    stats:
+      title: Candidate quality
+      source: candidates.quality
+    table:
+      title: Candidates by score
+      source: candidates.quality
+      limit: 10
+    list:
+      title: Strongest candidates
+      source: candidates.quality
+      limit: 5
+    progress:
+      title: Candidate quality
+      source: candidates.quality
+    insight:
+      title: Candidate quality
+      source: candidates.quality
+---
+
+# Candidate Analysis
+
+## Purpose
+
+- Report how strong the applicant pool is.
+- Report how much of it anyone has actually looked at, beside the quality figure.
+- Rank candidates by score so a shortlist has a starting point.
+
+## Capabilities
+
+- Summarize pool size, screening coverage, average score and interview count.
+- List or tabulate candidates by score.
+- Break the scored pool into quality bands.
+
+## Data
+
+Reads `candidates.quality`, which counts `JobApplication` records and their
+`ai_score`, joined to `AIInterview` records for interview coverage.
+
+## Analysis
+
+Coverage is reported next to quality, always. An average score computed from a
+fifth of the pool is not the pool's average, and reporting the first without the
+second is how a hiring dashboard talks itself into confidence.
+
+Scored candidates are grouped into four bands — 80 and above, 70 to 79, 50 to 69,
+and below 50 — because a mean hides whether a pool is uniformly mediocre or
+split between strong and weak.
+
+## Output
+
+Pool size, share screened, average score across scored candidates only, and
+interview count. Then candidates ranked by score with their role and stage.
+
+## Limitations
+
+- Unscored candidates are excluded from the average rather than counted as zero.
+  They are reported separately as the unscreened share.
+- A score is an AI screening score, not an interview outcome or a hiring decision.
+- Filtering by period counts applications by when they were received, not by when
+  they were screened.

skills/candidate-search.md

@@ -0,0 +1,29 @@
+---
+id: candidate-search
+name: Candidate Search
+description: Help Owliver search and summarize candidates.
+pages:
+  - candidates
+  - candidates-analysis
+status: active
+actions:
+  - navigate_to_candidates
+---
+
+# Candidate Search
+
+## Purpose
+
+Read the candidate pipeline on the current page and answer questions about who
+is waiting, who is strongest, and where screening is incomplete.
+
+## Capabilities
+
+- Summarize the candidate pipeline.
+- Identify candidates waiting on a decision.
+- Surface unscored or incomplete records.
+- Compare candidates by screening score.
+
+## Actions
+
+- navigate_to_candidates

skills/create-position.md

@@ -0,0 +1,69 @@
+---
+id: create-position
+name: Create Position
+description: Create a position by answering a few questions in the chat.
+pages:
+  - positions
+status: active
+prompt: Create a position
+triggers:
+  - create a position
+  - create position
+  # A client is the company a position is staffed for, so asking for one starts
+  # the same conversation — it simply leads with the company question.
+  - create a client
+  - create client
+  - add a client
+  - new client
+  - create a * position
+  - create * position
+  - new position
+  - new * position
+  - post a job
+  - post a * job
+  - open a role
+  - open a * role
+  - add a position
+  - i want to hire
+actions:
+  - create_position
+---
+
+# Create Position
+
+## Purpose
+
+Create a position without leaving the Positions page. Owliver asks for what it
+does not already know, one question at a time, offers the answers as chips, then
+reads the whole thing back before anything is written.
+
+No form opens. No page is navigated to. The record created is the same
+`JobPosting` the manual form writes, through the same create action.
+
+## Capabilities
+
+- Understand requests to create positions.
+- Read the role, location, pay, experience, English level and certifications out
+  of a single sentence.
+- Ask only for what the request did not already answer.
+- Offer each answer as a suggestion, so the whole flow can be clicked.
+- Read the position back for confirmation before creating it.
+- Create the position on the page you are already on.
+
+## Conversation
+
+Each line is `field | question | suggestions | required?`. Suggestions beginning
+with `@` come from the application's own data, so a role category added in the
+form is offered here without this file changing.
+
+- company | Which client is this role for? Type the company name. | | required
+- role_category | What role are you hiring for? | @roles | required
+- location | Where will this role be based? | Chennai; Bengaluru; Coimbatore; Bay Area; Other | required
+- pay | What is the pay range? | $18–$28/hr; $25–$35/hr; $30–$40/hr; Custom | required
+- min_experience_years | Any minimum experience? | No minimum; 1 year; 2 years; 3+ years | optional
+- english_required | What is the minimum English level? | @english | optional
+- certifications_required | Any required certifications? | @certifications; None | optional
+
+## Actions
+
+- create_position

skills/customer-service-training.md

@@ -0,0 +1,33 @@
+---
+id: customer-service-training
+name: Customer Service Training
+description: Training path for handling guests, complaints and recovery.
+skill: customer_service
+pages:
+  - positions
+  - candidates
+  - profile
+---
+
+# Customer Service Training
+
+Reading a guest, handling what goes wrong, and leaving them better than you
+found them. Three rungs: there is no fourth thing to be good at here, and
+inventing an Expert tier would make Expert mean less everywhere else.
+
+## Beginner
+
+Greet, read and serve a guest without supervision.
+
+## Intermediate
+
+Handle a complaint to resolution without escalating it.
+
+## Advanced
+
+Recover a badly broken experience and keep the guest.
+
+## Verification
+
+Owliver scores the response against the criteria on each module — what was
+acknowledged, what was offered, and whether the table was protected.

skills/executive-summary.md

@@ -0,0 +1,81 @@
+---
+id: executive-summary
+name: Executive Summary
+description: The whole workspace in one reading — positions, candidates, hires, talent and attendance.
+category: analytics
+pages:
+  - control-center
+status: active
+version: 1
+triggers:
+  - executive summary
+  - workspace summary
+  - overall summary
+  - state of the workspace
+  - brief me
+owliver:
+  enabled: true
+  suggestions:
+    - label: Give me an executive summary
+      capability: summary
+    - label: Show the headline figures
+      capability: stats
+  capabilities:
+    - summary
+    - stats
+    - card
+    - table
+    - insight
+  responses:
+    summary:
+      title: Workspace summary
+      source: workspace.summary
+    stats:
+      title: Workspace summary
+      source: workspace.summary
+    card:
+      title: Workspace summary
+      source: workspace.summary
+    table:
+      title: Workspace summary
+      source: workspace.summary
+    insight:
+      title: Workspace summary
+      source: workspace.summary
+---
+
+# Executive Summary
+
+## Purpose
+
+- Give a management-level reading of the whole workspace in one answer.
+- Draw every figure from the source that owns it, so the summary cannot drift
+  from the pages it summarizes.
+
+## Capabilities
+
+- Report open roles, candidates, hires, talent pool size, attendance and logged events.
+
+## Data
+
+Reads `workspace.summary`, which counts `JobPosting`, `JobApplication`, `Staff`,
+`WorkerProfile`, `UserActivity` and `ShiftRecord`.
+
+## Analysis
+
+Each figure is read from the domain that owns it rather than recomputed here. A
+domain with no records contributes a zero and says so in its detail line — the
+summary reports what is there, including an absence.
+
+## Output
+
+Six headline figures: open roles, candidates, hires, talent pool, attendance
+rate and events logged, each with a supporting detail.
+
+## Limitations
+
+- This is a count, not a diagnosis. Which figures are a problem is what Staffing
+  Risk, Operational Risk and Anomaly Detection answer.
+- Attendance is 0 where no shifts have been recorded, and the detail line says
+  so rather than implying nobody turned up.
+- No trend or comparison to a previous period is included.

skills/food-safety-training.md

@@ -0,0 +1,31 @@
+---
+id: food-safety-training
+name: Food Safety Training
+description: Training path for hazard awareness, temperature control and hygiene.
+skill: food_safety
+pages:
+  - positions
+  - candidates
+  - profile
+---
+
+# Food Safety Training
+
+Finding the hazard before it finds you. Every level is evidenced against a real
+prep station rather than a quiz score alone.
+
+## Beginner
+
+Identify the common hazards in a prep station.
+
+## Intermediate
+
+Hold, store and label food at safe temperatures through a full service.
+
+## Advanced
+
+Run a station to audit standard and correct others on it.
+
+## Verification
+
+Owliver judges the hazards identified and the ones missed.

skills/forge-skill-management.md

@@ -0,0 +1,64 @@
+---
+id: forge-skill-management
+name: Forge Skill Management
+description: Create workforce skills, training and verification in KROW Forge.
+pages:
+  - university
+status: active
+prompt: Create a skill
+triggers:
+  - create a skill
+  - create skill
+  - create a new skill
+  - create a skill for *
+  - create a * skill
+  - new skill
+  - add a skill
+  - build a skill
+  - draft a skill
+  - create a skill training
+  - create skill training
+  - add skill training
+  - create training
+  - create a training for *
+  - create training for *
+  - create a * training
+  - add training
+  - add training to *
+  - build training
+  - write training
+  - create a challenge for *
+  - add a challenge for *
+  - review the * skill
+  - review skill
+actions:
+  - open_create_skill_training
+  - open_create_training
+  - navigate_to_forge
+---
+
+# Forge Skill Management
+
+## Purpose
+
+Help an administrator build the workforce skill library: define what the
+workforce must be able to do, write the training that teaches it, decide what
+counts as proof, and say what the evaluation checks — using the Forge authoring
+flow the page already has.
+
+Owliver never publishes. It drafts, and hands the draft back to the existing
+flow for review, which is the same rule Create Position follows.
+
+## Capabilities
+
+- Create workforce skills
+- Create training
+- Define verification
+- Explain Forge skills
+- Navigate Forge workflows
+
+## Actions
+
+- open_create_skill_training
+- open_create_training
+- navigate_to_forge

skills/hiring-activity-assistant.md

@@ -0,0 +1,61 @@
+---
+id: hiring-activity-assistant
+name: Hiring Activity Assistant
+description: Answer questions about recent hiring activity on a position.
+pages:
+  - positions
+status: active
+triggers:
+  - hiring activity
+  - hiring summary
+  - hiring flow
+  - recent applications
+  - applications over time
+owliver:
+  enabled: true
+  # One suggestion per capability. A bare `- Show hiring activity` used to sit
+  # above these two: with no `capability:` it fell through to the first declared
+  # one — `summary` — so it and "Summarize hiring activity" were two chips for a
+  # single answer. Every suggestion here now names the capability it asks for,
+  # which is what makes a chip and an answer one-to-one.
+  suggestions:
+    - label: Summarize hiring activity for this position
+      capability: summary
+    - label: Show hiring activity as a flow
+      capability: flow
+  capabilities:
+    - summary
+    - flow
+  responses:
+    summary:
+      title: Hiring Activity Summary
+      source: position.activity
+      periods:
+        - today
+        - yesterday
+        - last-week
+    flow:
+      title: Hiring Activity Flow
+      source: position.activity
+      steps:
+        - today
+        - yesterday
+        - last-week
+---
+
+# Hiring Activity Assistant
+
+## Purpose
+
+Answer questions about how many people have applied to a position lately, in the
+panel beside the Positions experience.
+
+This is a separate definition from the Hiring Activity UI skill, and each is
+managed on its own list — but both name `position.activity`, so both are read by
+the one shared resolver from the same application records. Switching either off
+leaves the other exactly as it was.
+
+## Capabilities
+
+- Summarize applications to this position over today, yesterday and last week.
+- Draw the same counts as a flow inside the answer.

skills/hiring-history-analysis.md

@@ -0,0 +1,88 @@
+---
+id: hiring-history-analysis
+name: Hiring History Analysis
+description: Analyse completed hires — who was hired, how quickly, and how well they scored.
+category: hiring
+pages:
+  - hired-history
+status: active
+version: 1
+triggers:
+  - hiring history
+  - hire quality
+  - quality of hire
+  - time to hire
+  - who did we hire
+  - recent hires
+owliver:
+  enabled: true
+  suggestions:
+    - label: Who did we hire recently?
+      capability: list
+    - label: How is hiring performance?
+      capability: summary
+  capabilities:
+    - summary
+    - stats
+    - list
+    - table
+    - timeline
+    - insight
+  responses:
+    summary:
+      title: Hiring performance
+      source: hires.performance
+    stats:
+      title: Hiring performance
+      source: hires.performance
+    insight:
+      title: Hiring performance
+      source: hires.performance
+    list:
+      title: Recent hires
+      source: hires.recent
+      limit: 10
+    table:
+      title: Recent hires
+      source: hires.recent
+    timeline:
+      title: Recent hires
+      source: hires.recent
+---
+
+# Hiring History Analysis
+
+## Purpose
+
+- Report hires that have already happened.
+- Report how long they took and how well they scored.
+- Keep the record after the decision separate from the pipeline before it.
+
+## Capabilities
+
+- Summarize total hires, average days to hire, quality of hire and conversion rate.
+- List recent hires with their role and date.
+
+## Data
+
+Reads `hires.performance` and `hires.recent`, which join `Staff` to their
+`JobApplication` and `JobPosting` records.
+
+## Analysis
+
+Time to hire is the span between an application arriving and its final update.
+Quality of hire is the average AI score across scored applications. Conversion
+is hires as a share of all applications.
+
+## Output
+
+Headline hiring figures, and a list or timeline of recent hires.
+
+## Limitations
+
+- This is the record after the decision. Candidates still under consideration
+  are Candidate Analysis's question.
+- Time to hire is measured from the application record's timestamps, not from
+  when a role was opened.
+- Quality of hire is a screening score, not a performance review. Nothing here
+  reports how a hire has since worked out.

skills/hiring-pulse-analysis.md

@@ -0,0 +1,88 @@
+---
+id: hiring-pulse-analysis
+name: Hiring Pulse Analysis
+description: Read the recent rhythm of hiring — applications arriving, and how the funnel is converting.
+category: hiring
+pages:
+  - control-center
+  - analytics
+status: active
+version: 1
+triggers:
+  - hiring pulse
+  - hiring velocity
+  - hiring rhythm
+  - application rate
+owliver:
+  enabled: true
+  suggestions:
+    - label: What is the hiring pulse?
+      capability: summary
+    - label: Applications over recent periods
+      capability: flow
+  capabilities:
+    - summary
+    - flow
+    - stats
+    - insight
+  responses:
+    summary:
+      title: Hiring pulse
+      source: candidates.activity
+      periods:
+        - today
+        - last-7-days
+        - previous-month
+    flow:
+      title: Applications over time
+      source: candidates.activity
+      periods:
+        - today
+        - last-7-days
+        - this-month
+        - previous-month
+    stats:
+      title: Hiring performance
+      source: hires.performance
+    insight:
+      title: Hiring performance
+      source: hires.performance
+---
+
+# Hiring Pulse Analysis
+
+## Purpose
+
+- Report the recent rhythm of hiring: how many applications are arriving, and
+  how the funnel is converting them.
+- Distinguish a quiet week from a broken pipeline.
+
+## Capabilities
+
+- Count applications arriving across recent periods.
+- Report conversion, speed and quality of hire.
+
+## Data
+
+Reads `candidates.activity` for arrival counts over periods, and
+`hires.performance` for conversion, time-to-hire and quality.
+
+## Analysis
+
+Arrival counts are reported per period rather than as a single rate, because a
+rate averages away the shape — thirty applications in a month is a different
+situation depending on whether they arrived steadily or all on one day.
+
+## Output
+
+Applications per period, and headline conversion, speed and quality figures.
+
+## Limitations
+
+- **This is the analysis definition only.** The Hiring Pulse card that appears
+  on Krow pages is a separate UI configuration with its own placement and
+  period settings; the two are deliberately not the same definition and changing
+  one does not change the other.
+- Arrival counts are by application creation date, not by when a role opened.
+- A period with no applications reports zero, which is a real reading — it does
+  not distinguish "nobody applied" from "the role was not advertised".

skills/leadership-training.md

@@ -0,0 +1,37 @@
+---
+id: leadership-training
+name: Leadership Training
+description: Training path for briefing, assigning and correcting a floor team.
+skill: leadership
+pages:
+  - profile
+---
+
+# Leadership Training
+
+Pre-shift briefings, section assignments, and correcting a teammate without
+deflating them.
+
+Attached to Profile only: this is a development path an employee plans for
+themselves. It is deliberately not surfaced on the recruiter-facing pages, which
+demonstrates that `pages` controls visibility per page rather than globally.
+
+## Beginner
+
+Run a pre-shift briefing for a small section.
+
+## Intermediate
+
+Assign sections and hold a service to time.
+
+## Advanced
+
+Lead a floor team of eight through a large event.
+
+## Expert
+
+Build the rota and develop the leads who run it.
+
+## Verification
+
+Owliver evaluates the recorded briefing against ownership, timing and tone.

skills/learning-analysis.md

@@ -0,0 +1,80 @@
+---
+id: learning-analysis
+name: Learning Analysis
+description: Report what the training library holds and how far the workforce has progressed through it.
+category: workforce
+pages:
+  - krow-forge
+status: active
+version: 1
+triggers:
+  - learning analysis
+  - training progress
+  - course progress
+  - training library
+  - what training
+owliver:
+  enabled: true
+  suggestions:
+    - label: How is training progressing?
+      capability: progress
+    - label: What does the library hold?
+      capability: list
+  capabilities:
+    - summary
+    - progress
+    - list
+    - table
+    - stats
+  responses:
+    summary:
+      title: Training progress
+      source: workforce.training
+    progress:
+      title: Training progress
+      source: workforce.training
+    list:
+      title: Training paths
+      source: workforce.training
+    table:
+      title: Training paths
+      source: workforce.training
+    stats:
+      title: Training progress
+      source: workforce.training
+---
+
+# Learning Analysis
+
+## Purpose
+
+- Report what the training library holds.
+- Report how far the workforce has progressed against it.
+
+## Capabilities
+
+- Summarize training paths and progress against them.
+- List or tabulate the paths in the library.
+
+## Data
+
+Reads `workforce.training`, the existing source behind the Forge progression
+views.
+
+## Analysis
+
+Progress is counted against the paths the library actually defines, so adding a
+path changes the denominator rather than being reported as a sudden fall in
+completion.
+
+## Output
+
+Training paths with progress against each.
+
+## Limitations
+
+- A skill in Forge is something a person learns and is verified in. It is not an
+  Owliver capability, and the two must not be described as the same thing.
+- Progress is personal to the signed-in worker where their record is loaded, and
+  library-level otherwise.
+- This reports progression, not whether the training is any good.

skills/operational-risk.md

@@ -0,0 +1,94 @@
+---
+id: operational-risk
+name: Operational Risk
+description: Surface what is going wrong operationally across hiring and the roster.
+category: operations
+pages:
+  - control-center
+  - activity
+status: active
+version: 1
+triggers:
+  - operational risk
+  - operations risk
+  - what is going wrong
+  - backlog
+  - decisions owed
+owliver:
+  enabled: true
+  suggestions:
+    - label: What is going wrong operationally?
+      capability: list
+    - label: Summarize operational risk
+      capability: summary
+  capabilities:
+    - summary
+    - list
+    - table
+    - stats
+    - insight
+  responses:
+    summary:
+      title: Operational risk
+      source: operations.risk
+    list:
+      title: Operational risks
+      source: operations.risk
+    table:
+      title: Operational risks
+      source: operations.risk
+    stats:
+      title: Operational risk
+      source: operations.risk
+    insight:
+      title: Biggest operational risk
+      source: operations.risk
+---
+
+# Operational Risk
+
+## Purpose
+
+- Surface the operational problems that need somebody to act.
+- Draw them from every domain, because they feel like one problem to the person
+  who has to fix them.
+- Report nothing when the operation is running.
+
+## Capabilities
+
+- Detect strong candidates left awaiting a decision.
+- Detect an unscreened application backlog.
+- Detect open roles with no applicants.
+- Detect shifts going unworked.
+
+## Data
+
+Reads `operations.risk`, which joins `JobApplication`, `JobPosting` and
+`ShiftRecord`.
+
+## Analysis
+
+Four checks, each with a floor so ordinary operation does not trip them:
+
+1. **Decisions owed** — a candidate scoring 70 or above, screened or
+   shortlisted, and not moved on. Any such candidate counts.
+2. **Unscreened backlog** — three or more applications with no score.
+3. **Roles with no applicants** — any open role nobody has applied to.
+4. **Shifts unworked** — two or more absences or no-shows in the last 7 days.
+
+Findings are ordered most severe first. A finding is included only when it
+exists; an empty list means the operation is running, not that the check was
+skipped.
+
+## Output
+
+A count of open risks, then a row per risk naming what is wrong and the figures
+behind it.
+
+## Limitations
+
+- Thresholds are fixed rather than tuned to this workspace's volume.
+- "Decisions owed" assumes a screened, strong candidate should be progressed.
+  A candidate deliberately held is indistinguishable from one overlooked.
+- The shift check covers the last 7 days only; a longer pattern is Attendance
+  Analysis's question.

skills/overtime-analysis.md

@@ -0,0 +1,98 @@
+---
+id: overtime-analysis
+name: Overtime Analysis
+description: Analyse overtime hours, who is carrying them, and whether they are growing.
+category: workforce
+pages:
+  - analytics
+  - control-center
+status: active
+version: 1
+triggers:
+  - overtime
+  - extra hours
+  - hours worked
+  - working late
+owliver:
+  enabled: true
+  suggestions:
+    - label: How much overtime are we running?
+      capability: summary
+    - label: Who is working the most overtime?
+      capability: table
+    - label: Overtime over recent periods
+      capability: flow
+  capabilities:
+    - summary
+    - stats
+    - table
+    - progress
+    - flow
+    - insight
+  responses:
+    summary:
+      title: Overtime
+      source: workforce.overtime
+    stats:
+      title: Overtime
+      source: workforce.overtime
+    table:
+      title: Overtime by person
+      source: workforce.overtime
+    progress:
+      title: Overtime by person
+      source: workforce.overtime
+    flow:
+      title: Overtime over time
+      source: workforce.overtime
+      periods:
+        - last-7-days
+        - this-month
+        - previous-month
+    insight:
+      title: Overtime
+      source: workforce.overtime
+---
+
+# Overtime Analysis
+
+## Purpose
+
+- Report how much overtime the workforce is carrying.
+- Say who is carrying it, since a total spread evenly and a total sitting on one
+  person are different problems.
+- Show whether it is growing.
+
+## Capabilities
+
+- Summarize overtime hours and their share of scheduled time.
+- Compare overtime per person, most hours first.
+- Show overtime across recent periods.
+
+## Data
+
+Reads `workforce.overtime`, which compares scheduled hours to hours actually
+worked across `ShiftRecord` entries.
+
+## Analysis
+
+Overtime is reported both as hours and as a share of scheduled time. The ratio
+is what makes two teams comparable — forty hours means one thing across a
+fortnight and another across a year.
+
+Attendance and overtime are kept separate because one hides the other: a team
+can have perfect attendance and be running on thirty hours of overtime a week,
+and a single "workforce hours" figure would report that as healthy.
+
+## Output
+
+Total overtime hours, share of scheduled time, average per shift, and a row per
+person. Over periods, hours per window.
+
+## Limitations
+
+- Overtime is derived from recorded shift end times, not from an approvals
+  process. A shift that ran long appears here whether or not it was authorised.
+- A person with no overtime appears with zero, which is a real reading rather
+  than missing data.
+- Cost is not calculated. No pay rate is attached to a shift record.

skills/server-training.md

@@ -0,0 +1,37 @@
+---
+id: server-training
+name: Server Training
+description: Training path for server employees.
+skill: server
+pages:
+  - positions
+  - candidates
+  - profile
+---
+
+# Server Training
+
+The path from a first shift on the floor to running a section in a fine dining
+room. Each level is held only when every module on that rung has been completed
+and Owliver has passed the evidence.
+
+## Beginner
+
+Complete the fundamentals of guest service.
+
+## Intermediate
+
+Complete advanced table service and order management.
+
+## Advanced
+
+Complete fine dining service and guest recovery.
+
+## Expert
+
+Lead a floor team through a full service without supervision.
+
+## Verification
+
+Owliver evaluates the employee's practical response against the criteria on each
+module, and the level is awarded only on a pass.

skills/staffing-risk.md

@@ -0,0 +1,99 @@
+---
+id: staffing-risk
+name: Staffing Risk
+description: Identify open roles that will not fill on their own, and say why.
+category: workforce
+pages:
+  - positions
+  - control-center
+status: active
+version: 1
+triggers:
+  - staffing risk
+  - staffing gap
+  # `*` stands for anything in between, so one line covers "roles at risk",
+  # "roles that are at risk" and "roles most at risk" without listing each.
+  - roles*at risk
+  - positions*at risk
+  - understaffed
+owliver:
+  enabled: true
+  suggestions:
+    - label: Which roles are at risk?
+      capability: list
+    - label: Summarize staffing risk
+      capability: summary
+  capabilities:
+    - summary
+    - list
+    - table
+    - insight
+  responses:
+    summary:
+      title: Staffing risk
+      source: positions.risk
+    list:
+      title: Roles at risk
+      source: positions.risk
+      limit: 5
+    table:
+      title: Roles at risk
+      source: positions.risk
+    insight:
+      title: Biggest staffing risk
+      source: positions.risk
+---
+
+# Staffing Risk
+
+## Purpose
+
+- Name the open roles that are not going to fill without intervention.
+- Say which of four distinct problems each one has, because each has a
+  different fix.
+- Rank them so the reader knows which to deal with first.
+
+## Capabilities
+
+- Count the open roles currently at risk.
+- List those roles, worst first, with the reason for each.
+- Identify the single role most in need of attention.
+
+## Data
+
+Reads `positions.risk`, which joins open `JobPosting` records to their
+`JobApplication` records through the same `buildPosition` reading the Positions
+page renders from. No separate calculation, so a risk reported here and a health
+badge shown there cannot disagree.
+
+## Analysis
+
+A role is at risk when any of the following is true. They are kept apart rather
+than combined into a score, because the score would hide the only part that
+tells the reader what to do:
+
+1. **No applicants yet** — nobody has applied. Needs sourcing.
+2. **No candidate scoring 70 or above** — people applied, none are viable.
+   Needs the requirements or the pay revisiting.
+3. **Three or more unscreened** — a backlog nobody has looked at. Needs
+   screening.
+4. **Someone awaiting a decision** — a strong candidate has been screened and
+   not moved on. Needs a person to decide.
+
+Ranking weights an empty pipeline above a busy one that needs attention, since
+an empty pipeline takes longest to recover.
+
+## Output
+
+A count of roles at risk, then a row per role naming its department and its
+reasons. A role with no problems is not listed.
+
+## Limitations
+
+- Only open roles are considered. A paused or closed role is not at risk.
+- "Viable" means an AI score of 70 or above. An unscored candidate is not
+  counted as viable, so a role whose applicants nobody has screened will report
+  both an unscreened backlog and no viable candidate — those are two true
+  statements about the same cause.
+- This reads the pipeline, not the roster. It does not know how many people a
+  role needs, because no position in this workspace states a headcount.

skills/talent-pool-analysis.md

@@ -0,0 +1,93 @@
+---
+id: talent-pool-analysis
+name: Talent Pool Analysis
+description: Analyse the talent already known to this workspace — who is in it, how they score, and who is available.
+category: hiring
+pages:
+  - talent-pool
+status: active
+version: 1
+triggers:
+  - talent pool
+  - pool health
+  - available talent
+  - who is available
+  - supply of talent
+owliver:
+  enabled: true
+  suggestions:
+    - label: How healthy is the talent pool?
+      capability: summary
+    - label: Who is in the pool?
+      capability: table
+    - label: Strongest people in the pool
+      capability: list
+  capabilities:
+    - summary
+    - stats
+    - table
+    - list
+    - progress
+    - insight
+  responses:
+    summary:
+      title: Talent pool
+      source: talent.pool
+    stats:
+      title: Talent pool
+      source: talent.pool
+    table:
+      title: Talent pool
+      source: talent.pool
+    list:
+      title: Strongest in the pool
+      source: talent.pool
+      limit: 5
+    progress:
+      title: Talent pool
+      source: talent.pool
+    insight:
+      title: Talent pool
+      source: talent.pool
+---
+
+# Talent Pool Analysis
+
+## Purpose
+
+- Report who this workspace already knows and could place.
+- Say how much of the pool has been assessed, and how much has not.
+- Report availability and certification coverage.
+
+## Capabilities
+
+- Summarize pool size, how many are scored, and average score.
+- List or tabulate people by score.
+- Report how many have availability on file and how many are certified.
+
+## Data
+
+Reads `talent.pool`, which counts `WorkerProfile` records — their `krow_score`,
+`availability`, `certifications` and stated role.
+
+## Analysis
+
+The average score is computed across **scored profiles only**. Counting an
+unassessed profile as zero would report a healthy pool as poor in exact
+proportion to how much of it nobody has got to yet — a figure that gets worse
+as the pool grows, which is the opposite of what it should do.
+
+An unscored person is reported as "Not yet scored" rather than shown with a
+zero, because a zero reads as a bad assessment rather than an absent one.
+
+## Output
+
+Pool size, how many are scored and how many are not, average score across the
+scored, availability and certification counts, then people ranked by score.
+
+## Limitations
+
+- This is supply, not applicants. Someone in the pool has not applied to anything
+  by being here, and must not be described as a candidate for a role.
+- Availability is what a person stated on their profile, not a live calendar.
+- A profile with no score is unassessed, which is not the same as being weak.

skills/workforce-analytics.md

@@ -0,0 +1,88 @@
+---
+id: workforce-analytics
+name: Workforce Analytics
+description: Report workforce coverage — open roles and who has actually been hired into them.
+category: analytics
+pages:
+  - analytics
+status: active
+version: 1
+triggers:
+  - workforce analytics
+  - workforce coverage
+  - roles covered
+  - coverage
+owliver:
+  enabled: true
+  suggestions:
+    - label: How well are open roles covered?
+      capability: summary
+    - label: Coverage by role
+      capability: table
+  capabilities:
+    - summary
+    - stats
+    - table
+    - list
+    - progress
+    - insight
+  responses:
+    summary:
+      title: Workforce coverage
+      source: workforce.coverage
+    stats:
+      title: Workforce coverage
+      source: workforce.coverage
+    table:
+      title: Coverage by role
+      source: workforce.coverage
+    list:
+      title: Coverage by role
+      source: workforce.coverage
+    progress:
+      title: Coverage by role
+      source: workforce.coverage
+    insight:
+      title: Workforce coverage
+      source: workforce.coverage
+---
+
+# Workforce Analytics
+
+## Purpose
+
+- Report how many open roles have somebody hired into them.
+- Name the roles that have nobody yet.
+- State plainly where a role has not said how many people it needs.
+
+## Capabilities
+
+- Count open roles, those with someone hired, and those with nobody.
+- Report coverage per role.
+
+## Data
+
+Reads `workforce.coverage`, which reads open `JobPosting` records through
+`demandFor` — the same reading the Positions page uses — joined to `Staff` by
+`job_posting_id`.
+
+## Analysis
+
+A role's coverage is the number of people hired into it against the number it
+asked for. Where a role has not declared a headcount, this reports the hires and
+says the target is unstated. It does **not** assume one person per role: that
+would produce a confident fill percentage that means nothing, and nothing on
+screen would say so.
+
+## Output
+
+Counts of open roles, covered roles and uncovered roles, how many roles state a
+headcount, and a row per role.
+
+## Limitations
+
+- No position in this workspace currently states a headcount, so no fill
+  percentage is reported. The count of hires per role is real.
+- Only open roles are counted. Paused and closed roles are excluded.
+- Assignment records would refine this, but none exist yet; coverage is
+  therefore counted from hires.