agentleFS
MicSmMicSm/boffin/.cursor/rules/boffin-post-change-audit.mdc

boffin / rules

Cursor rule38 starsChanged 3 months agoMITView on GitHub

Pre-flight review and post-change kernel audit

The file

We split this boffin-post-change-audit.mdc into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If a change touches ownership, lifecycle, completion semantics, state-machine transitions, concurrent access, or logic currently encoded …

    line 20
  2. Distinctive:

    MUST verify the edit with an external check (the narrowest test or lint that proves it), not by re-reading your own reasoning or the diff …

    line 29
  3. Distinctive:

    MUST review the change as if a different author wrote it: read the final file region, not only the diff, and ask what a fresh reviewer …

    line 30
  4. Very specific:

    MUST re-read the relevant loaded leaf pack(s) and confirm no ## EXCLUDES pattern was introduced and no earlier-stage (S00-S03) invariant …

    line 31
  5. Very specific:

    MUST NOT silently drop a ledger finding: it ends as done (with its check) or skip:<reason>.

    line 32
  6. Distinctive:

    MUST present the filled ledger as the result.

    line 33

≈ “Common” is an estimate: at least 1 in 5 of the 6,668 Cursor rule files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
boffin-post-change-audit.mdc
---
description: Pre-flight review and post-change kernel audit
alwaysApply: true
---

# Pre-Flight Review and Post-Change Audit

## Before presenting code-task results

- Re-read the loaded pack index and all loaded leaf packs; do not rely on memory.
- For a focused change, confirm you stayed within the requested scope and proved
  the change with a check; the ledger checks below apply to refactor/review tasks.
- For a refactor/review task, confirm the Pass 1 audit ledger lists a row for
  every in-scope stage; a stage with no row is an incomplete audit, not a clean
  file.
- Confirm every ledger row is `done` or `skip`, that each `skip` carries a
  reason, and that no row recorded as `todo` was downgraded to `skip:clean`.
- Confirm every fix named in Pass 1 reasoning has its own ledger row; a fix
  discussed but left unrecorded is an incomplete audit.
- If a change touches ownership, lifecycle, completion semantics,
  state-machine transitions, concurrent access, or logic currently encoded
  through flags or sentinels, name the exact invariant from stages S01-S03 it
  must preserve before editing.
- On user request, render the ledger and the blocking earlier-stage invariants
  as concise ordered bullets in the user's language.

## After any code edit

- MUST verify the edit with an external check (the narrowest test or lint that proves it), not by re-reading your own reasoning or the diff alone.
- MUST review the change as if a different author wrote it: read the final file region, not only the diff, and ask what a fresh reviewer would flag.
- MUST re-read the relevant loaded leaf pack(s) and confirm no `## EXCLUDES` pattern was introduced and no earlier-stage (S00-S03) invariant was weakened to satisfy a later-stage (S04-S06) cleanup goal.
- MUST NOT silently drop a ledger finding: it ends as `done` (with its check) or `skip:<reason>`.
- MUST present the filled ledger as the result; do not add rows for kernels that found nothing, and never cite a kernel a row does not actually rest on.
- When a loaded kernel actually changed a code decision, leave one short host-language comment at that site: start it with the `boffin:` prefix and state, in the code's own vocabulary, what the code now does or refuses, never restating the kernel. The judgment verb is encouraged (`kept .../refused .../cut ...`); the `boffin:` prefix is the machine-readable key; the clause MUST NOT contain a kernel id, kernel scope-names, or any other pack vocabulary. Readability bar (the one mandatory wording check): mentally dropping the `boffin:` prefix, the remaining sentence MUST still read as a comment a senior engineer would write -- otherwise rewrite the clause or drop the mark; this is only a mental check, so always keep the `boffin:` prefix in the emitted comment. MUST NOT mark a purely stylistic rewrite of already-correct code that averts no failure and preserves no load-bearing invariant, nor emit anything when a `.boffin-trace-off` file exists at the repo root. Keep to about one mark per hunk and roughly three per changeset; no URLs, marketing, or tool jargon; only in venues that forbid tool marks, omit the `boffin:` prefix (keep the senior-voice clause alone).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.