feat(github-body-field): rewrite one issue-body field out-of-context#412
Merged
Conversation
… loading the body into agent context
The security-sync skills currently update a single `### Field`
section of a tracker issue body by reading the full 10-15 KB body
into agent context, regex-editing one field, and writing the whole
body back. That spends ~5 K tokens per single-field flip, in
addition to looping reporter-supplied content (the most sensitive
content on the tracker) through the agent for no reason.
`tools/github-body-field/` is a stdlib-only Python tool that does
the read / parse / replace / push in a subprocess. The agent only
sees the diff summary on stdout. Subcommands:
body-field get <N> --field "<name>" → prints one value
body-field set <N> --field "<name>" \
--value "<v>" → rewrites in place
body-field list <N> → prints all headings
Idempotent: a `set` whose value already matches skips the push
(saves an API call + an audit-log entry). Parser is a small state
machine tracking fenced code blocks so a literal `### foo` inside
a shell snippet never false-matches as a heading.
Migrates one call site in this PR — security-cve-allocate Step 4
item 1 (set *CVE tool link*) — to prove the pattern. The remaining
call sites (Public advisory URL, Remediation developer, Short
public summary for publish, Affected versions, etc.) will follow
in separate PRs.
41 unit tests cover the parser edge cases (fenced headings,
duplicate headings, empty values, last-section padding, idempotent
rewrites) and the CLI orchestration (dry-run, no-op short-circuit,
exit-code contract).
5 tasks
potiuk
added a commit
that referenced
this pull request
May 31, 2026
…PATCH out of context (#424) Every status-update-emitting skill (sync, import, allocate, dedupe, fix) folds its update into one rollup comment per tracker via the recipe in `tools/github/status-rollup.md`. The recipe currently walks the agent through: fetch the comment body, concatenate `<old body>` + ruler + new entry, PATCH the comment. That loops the full rollup body — which grows monotonically on long-running trackers — through agent context on every sync pass. `tools/github-rollup/` is the same shape as PR #412's `github-body-field`: a stdlib + `gh` subprocess wrapper that keeps the body out of agent context. CLI: github-rollup append <N> --action "<label>" --entry-body "..." github-rollup list <N> github-rollup latest <N> `append` auto-detects whether the issue already has a rollup comment (via the marker prefix `<!-- airflow-s status rollup v`); creates one if not, PATCHes the existing one if yes. `--now ISO8601`, `--user @handle`, and `--dry-run` flags support deterministic replay tests and pre-flight checks. Parser is a small state machine that: - recognises the rollup marker prefix (forward-compatible with future v2+ bumps), - iterates `<details><summary>YYYY-MM-DD · @user · Action</summary>...</details>` entries in document order, - preserves entries whose summary doesn't follow the canonical shape (returned with blank fields so round-trip stays safe), - tolerates missing close tags + trailing whitespace. 35 unit tests cover the parser (every edge case from `status-rollup.md`'s spec — multi-entry, missing marker, non- canonical summary, missing close tag) and the CLI orchestration (append-existing, create-new, dry-run create vs append, user override, all the exit-code contract paths, list, latest). The migration of the 5+ call sites that currently reference `tools/github/status-rollup.md` Step 2a will follow in subsequent PRs once the tool's behaviour is verified in practice. Adds workspace member + docs/labels-and-capabilities row in lock-step (caught by `check-workspace-members` prek hook).
potiuk
added a commit
to potiuk/magpie
that referenced
this pull request
Jun 1, 2026
…ity-suite refactor patterns Adds `optimize-skill` (capability:setup) — the refactoring sibling of `write-skill`. It takes an existing framework skill (or sweeps a set) and applies the five restructuring patterns proven on the security suite, as behavior-preserving proposals gated by the validator (green-before / green-after): - split — slim an oversized SKILL.md into linked siblings (the apache#410 pattern; addresses the PRINCIPLES.md P14 cap) - config-lift — move concrete values into <project-config> (apache#386/apache#387/apache#388) - out-of-context — read/PATCH one field without loading the body (apache#412 github-body-field, apache#424 github-rollup) - fetch-upfront — batch per-item round-trips (apache#347) - preflight-classifier — skip obvious no-ops before LLM passes (apache#414/apache#416) SKILL.md is 297 lines; the pass catalogue (smell / exemplar PR / mechanics / behavior-preservation guarantee / validation) lives in the patterns.md sibling. Reads only framework-internal files, so no injection-guard / Privacy-LLM callouts. Ships a step-diagnose eval (5 auto-comparable cases incl. an injection-resistance case) so the skill is not released without an eval (P8). Wires the skill into the capability->skill map and the eval index. Generated-by: Claude Code (Opus 4.8)
potiuk
added a commit
to potiuk/magpie
that referenced
this pull request
Jun 1, 2026
…ity-suite refactor patterns Adds `optimize-skill` (capability:setup) — the refactoring sibling of `write-skill`. It takes an existing framework skill (or sweeps a set) and applies the five restructuring patterns proven on the security suite, as behavior-preserving proposals gated by the validator (green-before / green-after): - split — slim an oversized SKILL.md into linked siblings (the apache#410 pattern; addresses the PRINCIPLES.md P14 cap) - config-lift — move concrete values into <project-config> (apache#386/apache#387/apache#388) - out-of-context — read/PATCH one field without loading the body (apache#412 github-body-field, apache#424 github-rollup) - fetch-upfront — batch per-item round-trips (apache#347) - preflight-classifier — skip obvious no-ops before LLM passes (apache#414/apache#416) SKILL.md is 297 lines; the pass catalogue (smell / exemplar PR / mechanics / behavior-preservation guarantee / validation) lives in the patterns.md sibling. Reads only framework-internal files, so no injection-guard / Privacy-LLM callouts. Ships a step-diagnose eval (5 auto-comparable cases incl. an injection-resistance case) so the skill is not released without an eval (P8). Wires the skill into the capability->skill map and the eval index. Generated-by: Claude Code (Opus 4.8)
potiuk
added a commit
that referenced
this pull request
Jun 1, 2026
…ity-suite refactor patterns (#427) Adds `optimize-skill` (capability:setup) — the refactoring sibling of `write-skill`. It takes an existing framework skill (or sweeps a set) and applies the five restructuring patterns proven on the security suite, as behavior-preserving proposals gated by the validator (green-before / green-after): - split — slim an oversized SKILL.md into linked siblings (the #410 pattern; addresses the PRINCIPLES.md P14 cap) - config-lift — move concrete values into <project-config> (#386/#387/#388) - out-of-context — read/PATCH one field without loading the body (#412 github-body-field, #424 github-rollup) - fetch-upfront — batch per-item round-trips (#347) - preflight-classifier — skip obvious no-ops before LLM passes (#414/#416) SKILL.md is 297 lines; the pass catalogue (smell / exemplar PR / mechanics / behavior-preservation guarantee / validation) lives in the patterns.md sibling. Reads only framework-internal files, so no injection-guard / Privacy-LLM callouts. Ships a step-diagnose eval (5 auto-comparable cases incl. an injection-resistance case) so the skill is not released without an eval (P8). Wires the skill into the capability->skill map and the eval index. Generated-by: Claude Code (Opus 4.8)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
tools/github-body-field/Python tool that reads, parses, and rewrites a single### Fieldsection of a GitHub issue body in a subprocess — the body never crosses into agent context.apply-and-push.md/signals-to-actions.mdwill follow in separate PRs.CLI:
Idempotent — a
setwhose value already matches the current is a no-op (no API write, no audit-log entry). Parser tracks fenced code blocks so a literal### fooinside a shell snippet never false-matches as a heading.Test plan
ruff check/ruff format/mypy— greenprek run --files <staged>— greenskill-and-tool-validate— greenlycheeon the new README — clean🤖 Generated with Claude Code