stale-brain
VerifiedAudit AI agent memory and instruction files: CLAUDE.md, AGENTS.md, .cursorrules, GEMINI.md, copilot-instructions.md and every other major agent's memory files (13 agents, 18+ file locations) against the live repository. Use when the user runs /stale-brain or asks to audit, verify, refresh, lint, or fix agent memory, rules, or instruction files; when they complain the agent "ignores rules", "keeps getting it wrong", or that instructions "feel outdated"; and mid-task, whenever an instruction in a loaded memory file contradicts observed reality (a referenced path is missing, a named script doesn't exist, the stated package manager doesn't match the lockfile). Extracts checkable claims, verifies each with cited evidence, ages the rest with half-lives, meters token cost, and proposes approve-only fixes.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 10.6 KB |
| assets/demo.svg | other | 4.4 KB |
| assets/hero.svg | other | 1.1 KB |
| assets/pulse.svg | other | 3.9 KB |
| references/claim-types.md | markdown | 7.9 KB |
| references/memory-sources.md | markdown | 3.9 KB |
| references/output-format.md | markdown | 15.0 KB |
references/memory-sources.md
View on GitHub# Memory sources: where every agent keeps its brain
stale-brain audits *agent-facing* instruction and memory files. This map is the
inventory checklist for step 1. Glob for all of them; report only what exists.
## Source map
| Agent / tool | Files (repo-relative unless noted) | Loading |
|---|---|---|
| Claude Code | `CLAUDE.md` (root **and** any subdirectory), `CLAUDE.local.md`, `.claude/**/*.md` (memory, rules, decision records), `~/.claude/CLAUDE.md` (user-global; audit only if user opts in) | always-loaded (root chain); `.claude/` memory varies |
| OpenAI Codex / cross-tool standard | `AGENTS.md` (root and any subdirectory) | always-loaded |
| Cursor | `.cursorrules` (legacy), `.cursor/rules/**/*.mdc` | always or glob-scoped (check each rule's frontmatter `globs`/`alwaysApply`) |
| GitHub Copilot | `.github/copilot-instructions.md`, `.github/instructions/*.instructions.md` | always-loaded / glob-scoped |
| Gemini CLI | `GEMINI.md` (root and subdirectories) | always-loaded |
| Windsurf | `.windsurfrules`, `.windsurf/rules/**` | always-loaded |
| Cline | `.clinerules` (file or directory `.clinerules/**`) | always-loaded |
| Roo Code | `.roo/rules/**` | always-loaded |
| Aider | `CONVENTIONS.md` (when referenced from `.aider.conf.yml`; check the conf) | opt-in |
| Zed | `.rules` | always-loaded |
| Amazon Q Developer | `.amazonq/rules/**` | always-loaded |
| JetBrains Junie | `.junie/guidelines.md` | always-loaded |
| OpenHands | `.openhands/microagents/repo.md`, other `.openhands/microagents/*.md` | always / trigger-scoped |
| Generic / project-specific | any file the user names ("also audit docs/agent-notes.md") | as stated |
Unknown agents appear constantly. The fallback heuristic: a markdown/text file whose
audience is *the agent* (imperative instructions about the repo) is in scope if the user
confirms it; a file whose audience is *humans* is not.
## Always-loaded vs on-demand
This classification drives the token meter. **Always-loaded** files are injected into
every session; their tokens are a per-session tax, and stale claims in them do
per-session damage. **On-demand** files (glob-scoped rules, opt-in docs) cost tokens only
when triggered; weight them accordingly and say so in the meter.
When loading behavior is ambiguous (e.g. a `.cursor/rules/*.mdc` without frontmatter),
classify as always-loaded (the conservative direction for a cost meter) and mark it
with `†` in the inventory table.
## Deliberately out of scope (unless the user asks)
- `README.md`, `CONTRIBUTING.md`, `docs/**`: human documentation. Auditing docs is a
different (bigger) job; naming these in the report as "not audited" is fine, judging
them is not.
- Lockfiles, manifests, CI configs: these are *evidence*, not memory. They are what
claims get checked against.
- Prompt/transcript archives (`.specstory/`, session logs): history, not instructions.
## Cross-file conflict rule
During extraction, index claims by topic (package manager, test command, directory
layout, owner). Two memory files asserting different values for the same topic is a
conflict finding. Agents that read both files are being told two different truths; that
is worse than one stale file. Mechanics:
- **IDs and anchors.** Conflicts are numbered `X1…Xn` (separate from claim IDs) and
anchored to both sides: `CLAUDE.md:12 ↔ .cursorrules:3`.
- **Not counted.** X-findings are excluded from the claim counts and the health score;
the underlying claims are already scored individually.
- **Folding.** When verification already flags one or both sides (🔴/🟡), fold the
conflict into those findings' evidence lines ("also conflicts with .cursorrules:3")
instead of rendering a third finding. Render a standalone X-finding only when neither
side is individually flagged (typically both ⚪: two files disagree and the repo offers
no evidence either way); its action is "ask the user which is true", and it appears at
the end of the 🔴 section.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |