sermilion/bill-feature-verify

Feature Verify

Project Overrides

If .agents/skill-overrides.md exists in the project root and contains a ## bill-feature-verify section, read that section and apply it as the highest-priority instruction for this skill. The matching section may refine or replace parts of the default workflow below.

If an AGENTS.md file exists in the project root, apply it as project-wide guidance.

Precedence for this skill: matching .agents/skill-overrides.md section > AGENTS.md > built-in defaults. When you reuse another skill as part of this workflow, also apply that skill's matching override section when present.

Workflow Overview

Task Spec + PR → Extract Criteria → Gather Diff →
  Feature Flag Audit (if needed) →
  Code Review (dynamic agents) →
  Completeness Audit (criteria vs code) →
  Consolidated Verdict

Step 1: Collect Inputs

Ask the user for the task spec (paste, file path, directory) and PR (number, branch, or commit range).

Accept PDFs (read in page ranges if >10 pages), markdown, inline text. If total text exceeds ~8,000 words, ask which sections are most relevant.

Step 2: Extract Acceptance Criteria

After reading the spec, produce in one pass:

1. Acceptance criteria — numbered list
2. Non-goals — things explicitly out of scope
3. Rollout expectation — does the spec require guarded rollout?
4. Key technical constraints — specific patterns, APIs, or architectural requirements

Then ask: Confirm or adjust the criteria before I review the PR.

Step 3: Gather PR Diff

Based on user input, gather changes via gh pr diff, git diff, or git log.

Step 4: Feature Flag Audit (conditional)

Read audit-rubrics.md for the full feature flag audit rubric and output format.

Step 5: Code Review

Run bill-code-review against the PR diff. Follow the full skill instructions including any matching .agents/skill-overrides.md section.

Step 6: Completeness Audit

Read audit-rubrics.md for the completeness audit format and rules.

Step 7: Consolidated Verdict

Read audit-rubrics.md for the verdict format and PR comment instructions.

Telemetry

This skill emits skillbill_feature_verify_started and _finished events via the feature_verify_started / feature_verify_finished MCP tools.

For the shared telemetry contract including the orchestrated flag semantics, follow telemetry-contract.md.

Skill-specific telemetry fields

Standalone invocation:

1. Call feature_verify_started after Step 2 (criteria confirmed) with acceptance_criteria_count, rollout_relevant, and spec_summary. Save the returned session_id.
2. Call feature_verify_finished after Step 7 (verdict delivered) with session_id, feature_flag_audit_performed, review_iterations, audit_result (all_pass / had_gaps / skipped), completion_status (completed / abandoned_at_review / abandoned_at_audit / error), and optional gaps_found list.

Orchestrated invocation (when called from another workflow that passes orchestrated=true):

1. Skip feature_verify_started.
2. Call feature_verify_finished with orchestrated=true and all started+finished fields combined. The tool returns {"mode": "orchestrated", "telemetry_payload": {...}}.
3. Return that payload to the orchestrator — it will embed it in its own finished event.

Nested child tools: when this skill runs bill-code-review as part of the verify workflow, this skill is itself a parent. Pass orchestrated=true to import_review and triage_findings, collect the returned telemetry_payload, and include it in the gaps_found context or alongside the verify payload. (In practice, standalone feature-verify will let its own skillbill_feature_verify_finished event be the single emitted event; the nested review is parent-owned.)

Skills Reused

• bill-code-review — shared router for stack-specific code review
• bill-feature-guard — optional rollout checklist when the spec, diff, or repo policy requires it