Which AI agents does this loop work with?

Ralph the docs backlog is written for claude-code, codex.

Is this loop safe to run unattended?

This loop is graded B (75/100) by looprepo's automatic safety evaluator, which checks for exit conditions, iteration budgets, verification steps, and scoping before publishing.

Loop/ralph Docsmedium riskintermediatesafety B · 75 · open-ended Geoffrey Huntley

Ralph the docs backlog

Document one undocumented public module per fresh-context iteration, verifying every code sample compiles and accumulating style rules in guardrails so the docs read like one author wrote them.

prompt

→ Claude

/loop fresh context each iteration: read docs-backlog.json, docs/STYLE.md, and .ralph/guardrails.md; pick the top undocumented module, write its reference page with a runnable example, execute the example to prove it works, and mark the module done; add any style or structure decision to .ralph/guardrails.md; stop when the backlog is empty or after 20 turns

claude-code · codex

Implementation note

Executing every example is the anti-hallucination guarantee — docs written this way cannot describe an API that does not exist. Seed docs/STYLE.md with a page you already like.

Source: Geoffrey Huntley (Ralph technique) ↗

More docs loops

Keep docs in sync with main

Loop/schedule

On every push to main, check whether changed code drifted from the docs in /docs and open a PR fixing anything out of date.

prompt

→ Claude

/schedule on every push to main, check whether the changed code drifted from the docs in /docs, and open a PR fixing anything out of date

docshigh riskclaude-code codex

Set "remove every TODO comment in src/ and…

Loop/goal

Community goal loop for docs, sourced from github. Verified exit condition, evaluator-gated.

prompt

→ Claude

/goal set "remove every TODO comment in src/ and explain each removal" stop after 8 turns

docslow riskclaude-code

source ↗

README freshness check

Loop/schedulelooprepo

Once a week, verify every command, path, and badge in the README actually works against the current codebase and open a PR fixing anything stale.

prompt

→ Claude

/schedule every Friday at 3pm, verify the README: run each documented setup and usage command in a clean checkout, check that referenced files and scripts exist, and open a PR correcting anything that fails or has drifted, with a note explaining each fix

docsmedium riskclaude-code codex