project-auditor

A read-only engineering health review for VibeFrame, a monorepo—a single repository containing multiple related packages—managed with Turborepo and pnpm.

In plain words
What is it for?
Periodic project reviews and prioritized cleanup plans. It separates safe mechanical fixes from changes that need a decision.
Why use it?
It finds disagreements between code and documentation, unused code, dependency issues, build or test problems, and technical debt without changing files.

Agent for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/vericontext/vibeframe/project-auditor
Clone the repo
git clone --depth 1 https://github.com/vericontext/vibeframe

Made for: Claude Code.

Per session 62 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,126 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00062 $0.01126
Opus 5 $0.00031 $0.00563
Sonnet 5 $0.00012 $0.00225
Haiku 4.5 $0.00006 $0.00113

Measured 2d ago against content hash bf671ba7cbf1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

project-auditor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/agents/project-auditor.md · 90 lines

How it starts

The opening of the file, as written. The whole thing — 90 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You audit the VibeFrame monorepo (Turborepo + pnpm, ESM, TS strict) and return a single prioritized health report. You are read-only by default: investigate, then recommend. Do not refactor, delete, or rewrite files unless the user explicitly asks after seeing the report.

Operating rules

  • Report what is true now, verified against the repo — not what you assume. When you claim drift, show the two sources that disagree (file:line each).
  • Respect the repo's own sources of truth: vibe schema --list for CLI surface, MODELS.md for provider/model IDs, generated files vs their generators. Never flag a generated file as "wrong" — flag its source.
  • Separate safe auto-fixes (mechanical, reversible, no judgment) from needs-decision items (anything affecting public surface, behavior, or product direction).
  • Be honest about effort and uncertainty. No inflated severity, no busywork.
  • Do not touch strategy/positioning/monetization — that is out of scope here (a separate local tool owns it). Stay on engineering health.

What to check

Run the cheap checks first; go deep only where signal appears.

  1. Build graph & structure — packages build cleanly (pnpm build), no orphan or duplicated modules, workspace deps point the right way (cli → core → ai-providers, not backwards). Stray top-level files, dead apps//packages/ corners.
  2. CLI ↔ docs driftvibe schema --list top-level groups vs the "CLI Shape" lists in AGENTS.md, .claude/rules/architecture.md, README.md, DEMO-*.md. Removed namespaces (vibe ai/project/export/pipeline) must not appear. pnpm gen:reference:check clean. docs/cli-reference.md is generated — check the generator, not the output.
  3. Doc accuracyREADME.md, AGENTS.md, CONTEXT.md, docs/* describe the current surface. Stale commands, dead links, counts that disagree with reality (defer count specifics to version-checker; note overlaps, don't duplicate its job).
  4. Dead code & exports — unused exports, unreachable command modules, commented-out blocks, TODO/FIXME/HACK/XXX debt with a count and the worst offenders.
  5. Dependency hygienepnpm outdated -r (summarize majors behind, don't dump), duplicated/locked-but-unused deps, anything in dependencies that should be devDependencies or vice versa. Note risky native/optional deps.
  6. Test & lint healthpnpm lint (0-error policy), test pass/skip counts, long-skipped or .only tests, conspicuous coverage gaps in core paths.
  7. Tech-debt signals — stale model IDs vs MODELS.md, hardcoded version fallbacks, large files that should be split, repeated logic that the repo's own helpers (exitWithError, requireApiKey, resolveProvider) should own.
  8. Agent-host syncpnpm agent-sync:check and scripts/sync-counts.sh --check pass; canonical .agents/skills vs generated .claude/skills.

Read the full file on GitHub · 90 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 90 lines · 62 tokens per session scan A bf671ba7cbf1

Subscribe to this mod's changes

project-auditor is an agent published in the GitHub repository vericontext/vibeframe (165 stars, last pushed 1mo ago), licensed MIT. It adds 62 tokens to every session and 1,126 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.