detect-stale-documentation

A documentation-freshness checker for a knowledge index. It probes saved reference points and records whether each one is current, drifted, or unknown.

In plain words
What is it for?
Use it to run freshness checks, investigate drifted or unknown entries, and read the knowledge index.
Why use it?
It shows when stored guidance may no longer match its source, without treating uncertainty as a failure. The index must be updated separately after checking.

Skill for Claude CodeCodex

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 skills/everywan-dev/claude-code-engineering/detect-stale-documentation
Any agent
npx skills add everywan-dev/claude-code-engineering --skill detect-stale-documentation
Clone the repo
git clone --depth 1 https://github.com/everywan-dev/claude-code-engineering

Made for: Claude Code, Codex.

Per session 70 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 669 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 $0.00070 $0.00669
Opus 5 $0.00035 $0.00334
Sonnet 5 $0.00014 $0.00134
Haiku 4.5 $0.00007 $0.00067

Measured yesterday against content hash 3eb63f387b58, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

detect-stale-documentation 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 yesterday.

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.

skills/detect-stale-documentation/SKILL.md · 63 lines

How it starts

The opening of the file, as written. The whole thing — 63 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Detect stale documentation

Run the probes

python3 -m validated_memory probe

Runs every anchor of every active unit (one superseded within the validated set is excluded) through the probe registered for its kind in validated-memory.md's probes: map, and appends one record per anchor to verdicts.jsonl -- append-only, never rewritten. A summary goes to stdout:

probe: 3 anchor(s) probed across 1 unit(s): 1 current, 1 drifted, 1 unknown

A drifted or unknown verdict is data, not a finding: it never gates probe. Only a source-validation ERROR, or the log failing to write, does.

Read the verdicts

python3 -m validated_memory derive

probe only records; derive is what turns the recorded verdicts into the verdict column of knowledge-index.md. Run derive after probe to see the update -- the index does not refresh itself.

The ternary, fail-explicit

Every verdict is one of three values, and the framework never guesses a fourth:

  • current -- the probe re-checked the anchor and it still holds.
  • drifted -- the probe re-checked the anchor and it no longer holds.
  • unknown -- the probe could not tell, or the anchor was never probed at all. An anchor that has simply not been checked yet reads exactly the same as one a probe failed to resolve: fail-explicit means "we don't know" is never silently folded into "current."

A unit's verdict is the worst of its anchors' verdicts (drifted > unknown > current), read from the knowledge-index.md table. The exact cell format -- including how systems behind an unknown anchor are listed -- is derive's algorithm, documented once in the README's derive section; read the cell there, do not re-derive it by hand.

Register a probe for a new anchor kind

Add an entry to the probes: map in validated-memory.md: <kind>: <command>. The command is run without a shell, receiving the anchor's envelope as JSON on stdin and answering {"verdict": ..., "detail": ...} as JSON on stdout, exit 0 -- see the README's "Probe contract" (under probe) for the exact shape. The plugin ships one probe, git_ref (freshness of a git ref), already registered by init; see the README's "The bundled git_ref probe".

Read the full file on GitHub · 63 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. yesterday First seen · 63 lines · 70 tokens per session scan A 3eb63f387b58

Subscribe to this mod's changes

detect-stale-documentation is a skill published in the GitHub repository everywan-dev/claude-code-engineering (2 stars, last pushed 13d ago), licensed Apache-2.0. It adds 70 tokens to every session and 669 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 2026-08-31.