Help us improve
Share bugs, ideas, or general feedback.
From pharaoh
Orchestrates parallel full project audits by dispatching 5 atomic skills (coverage, lifecycle, standards, reviews, processes) to Papyrus workspace for deduplication. Emits aggregated findings for CI gates or pre-release reviews.
npx claudepluginhub useblocks/pharaoh --plugin pharaohHow this skill is triggered — by the user, by Claude, or both
Slash command
/pharaoh:pharaoh-audit-fanoutThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Invoke at the start of a full project audit, CI gate, or pre-release review. Produces a deduplicated, terminology-consistent findings report covering 5 sub-areas in parallel.
Adversarial review of code subsystems or non-code artifacts (design docs, plans, concepts) through parallel analytical lenses. Dispatches analysis agents across four tailored lenses and synthesizes findings.
Audits sphinx-needs projects for coverage gaps across 10 categories plus cross-artefact consistency, emitting a single prioritized gap report with recommended actions.
Runs a repo-wide governance audit (validators, cross-cutting checks, aggregate counters) across Claude Code and non-Claude AI clients. Detects runtime and dispatches to native sub-agent or executes inline audit flow.
Share bugs, ideas, or general feedback.
Invoke at the start of a full project audit, CI gate, or pre-release review. Produces a deduplicated, terminology-consistent findings report covering 5 sub-areas in parallel.
This is an orchestrator skill. Atomicity criteria (a) does not apply — the skill is compositional by construction. Its 5 component skills each pass (a)-(e) individually.
| # | Atomic skill | Target issues |
|---|---|---|
| 1 | pharaoh-coverage-gap | orphans, unverified req, uncovered req-verification |
| 2 | pharaoh-lifecycle-check | invalid state transitions, stale reviews |
| 3 | pharaoh-standard-conformance | ISO 26262 §6 / ASPICE indicator gaps |
| 4 | pharaoh-review-completeness | missing reviewer / approved_by fields |
| 5 | pharaoh-process-audit | duplicate req, contradictory pair, missing FMEA, broken back-link |
All 5 subagents write findings to the project's .papyrus/ workspace using pharaoh-finding-record. The deterministic-ID scheme in pharaoh-finding-record provides natural dedup via Papyrus's FileLock-guarded ID-collision semantics.
Subagents SHOULD NOT retry on action: duplicate — that's the intended signal that another subagent already covered this issue.
Ensure <project_dir>/.papyrus/ exists and is writable. If not, initialize via papyrus --workspace <project_dir>/.papyrus init (zero-config, silos preset).
In a production Pharaoh deployment, dispatch via Agent tool with subagent_type: general-purpose. In the Phase 4b harness, parallelism is achieved at the harness level via concurrent.futures.ThreadPoolExecutor over 5 separate claude -p invocations, each primed with one sub-area's task.
Each subagent receives:
pharaoh-finding-record for every findingOnce all 5 subagents complete, read the Papyrus workspace via:
papyrus --workspace <project_dir>/.papyrus recall --tag category:* --format full
Emit as the final audit report (JSON list of findings, ordered by category).
Input: {project_dir: path, scope?: "full"|"partial"}. Default scope is full (all 5 sub-areas). Partial scope limits to a subset of sub-areas (not used in Phase 4b testing).
Output: JSON list of aggregated findings, one object per (category, subject_id) pair:
[
{"category": "orphan_arch", "subject_id": "arch__orphan_0", "finding_text": "...", "reporter_id": "coverage-gap"},
...
]
.papyrus/ workspace uninitializable → fall back to sequential single-agent pharaoh-process-audit with stderr warning.1 subagent timeout → abort and surface partial Papyrus state.
Consumed by CI gates, on-demand user audits, and release governance workflows. Output is directly consumable by pharaoh-audit-report (future skill) or printed as-is.