reporting

Reporting

Invokable anytime via /geas:reporting. The orchestrating skill calls this at phase transitions, session start, and evolving phase entry.

Produces a markdown summary to the conversation and writes .geas/state/health-check.json.

---

Section 1: Debt Summary

Read .geas/missions/{mission_id}/evolution/debt-register.json.

Report the following counts:

By severity (use rollup_by_severity):

• critical, high, medium, low

By kind (use rollup_by_kind):

• code_quality, architecture, security, docs, ops, test_gap, product_gap

By status — count items in items[] by their status field:

• open, accepted, scheduled, resolved, dropped

If the file does not exist, report: "No debt register found — run after at least one phase completes."

---

Section 2: Gap Summary

Read the latest .geas/missions/{mission_id}/evolution/gap-assessment-*.json file (most recent by timestamp).

Report:

• Fully delivered: count + list of goal values where delivery_status = "fully_delivered"
• Partially delivered: count + for each item: goal, delivered portion, remaining portion
• Not delivered: count + for each item: goal + rationale

Compute gap closure ratio:

ratio = fully_delivered / (fully_delivered + partially_delivered + not_delivered)

Report as a percentage (e.g., 75%). If denominator is 0, report N/A.

If the file does not exist, report: "No gap assessment found — run after at least one phase completes."

---

Section 3: Health Signals

Calculate all 8 signals from protocol doc 12. For each signal: read the source data, compute the value, compare against the threshold, and determine whether it is triggered.

Signal	Threshold	Source	Calculation
memory_bloat	entries > 100 with 0 reuses in last 10 tasks	.geas/state/memory-index.json	Count index entries where reuse_count = 0
review_gap	> 20% miss rate in last 5 tasks	.geas/ledger/events.jsonl	Count tasks without specialist review evidence / total tasks (last 5)
gate_quality_issue	> 30% iterate rate in last 10 tasks	.geas/ledger/events.jsonl	Count final_verdict = iterate events / total final verdict events (last 10)
contradiction_accumulation	2+ stable memories with contradiction_count >= 3	.geas/state/memory-index.json	Count index entries where state = stable and contradiction_count >= 3
repeated_failure_class	same failure class 3+ times	.geas/ledger/events.jsonl	Group gate failure events by blocking_dimensions; flag any group with count >= 3
debt_stagnation	accepted >= 2x resolved per phase	.geas/missions/{mission_id}/evolution/debt-register.json	accepted_count / resolved_count (if resolved = 0 and accepted > 0, value = infinity → triggered)
scope_control_weakness	> 30% scope change after impl contract in last 5 tasks	.geas/ledger/events.jsonl	Count revalidation_triggered events after implementation_contract_approved per task (last 5); flag if > 30% of those tasks have at least one such event
worker_low_confidence	> 25% confidence <= 2 in last 10 tasks	.geas/missions/{mission_id}/tasks/*/worker-self-check.json	Count self-check files with confidence <= 2 / total self-check files (last 10 by created_at)

Mandatory Responses

When a signal is triggered, include the mandatory response in the output and take the described action:

1. memory_bloat — orchestration_authority batch-reviews memories with 0 reuses at the next retrospective. Items past their review_after date are transitioned to decayed.
2. review_gap — orchestration_authority activates blocking at the pre_gate hook for specialist review omissions starting from the next task.
3. gate_quality_issue — orchestration_authority creates a rule candidate for rubric criteria clarification. Adjusts evidence gate thresholds if needed.
4. contradiction_accumulation — immediately transition the affected memory entries to under_review and apply doc 08 decay rules.
5. repeated_failure_class — automatically register the failure pattern as a memory candidate and trigger the Retrospective to Rule Update process from doc 13.
6. debt_stagnation — orchestration_authority creates a debt resolution plan during phase review. Prioritize debt resolution tasks in the next phase scheduling.
7. scope_control_weakness — orchestration_authority strengthens the implementation contract approval process. Makes re-approval mandatory on scope changes.
8. worker_low_confidence — orchestration_authority reviews task granularity and improves L1/L2 memory quality in context packets.

Missing source data

If a source file does not exist, mark the signal as value: null, triggered: false and note the missing source in the summary.

---

Section 4: Output

Write .geas/state/health-check.json:

{
  "timestamp": "<ISO 8601>",
  "signals": [
    {
      "name": "memory_bloat",
      "value": 15,
      "threshold": 100,
      "triggered": false
    },
    {
      "name": "debt_stagnation",
      "value": 2.5,
      "threshold": 2.0,
      "triggered": true,
      "mandatory_response": "orchestration_authority creates a debt resolution plan during phase review. Prioritize debt resolution tasks in the next phase scheduling."
    }
  ]
}

Include all 8 signals in the signals array. Only include mandatory_response on entries where triggered: true.

Then print a markdown summary to the conversation:

## Health Report — <ISO 8601 timestamp>

### Debt
- Critical: N  High: N  Medium: N  Low: N
- Open: N  Accepted: N  Scheduled: N  Resolved: N  Dropped: N

### Gaps
- Fully delivered: N  Partially: N  Not delivered: N  (Closure ratio: X%)

### Health Signals
| Signal | Value | Threshold | Status |
|--------|-------|-----------|--------|
| memory_bloat | 15 | 100 | OK |
| debt_stagnation | 2.5 | 2.0 | TRIGGERED |
...

### Required Actions
- debt_stagnation: orchestration_authority creates a debt resolution plan during phase review.
...

If no signals are triggered, print: "All health signals within thresholds."

---

Section 5: Product Authority Briefing

The reporting skill also generates product authority briefings at key milestones (phase transitions, mission completion). When invoked at a milestone:

1. Summarize completed tasks, their verdicts, and any open debt
2. List key decisions made since the last briefing
3. Highlight risks or signals that require product authority attention
4. Write the briefing to .geas/missions/{mission_id}/phase-reviews/{phase}-briefing.md

---

Section 6: Session End Summary

At session end, generate an audit-trail summary of decisions, tasks, and agent telemetry:

1. Read .geas/ledger/events.jsonl for all events in the current session
2. Summarize: tasks started/completed, gate results, decisions made, escalations
3. Include agent spawn counts and which agents were most active
4. Write to .geas/summaries/session-{timestamp}.md