Conversation
Agent memory does not survive a new session or a new machine, so the hazards this convergence surfaced belong in the tree rather than in a local note. AGENTS.md gains a warning in its intro, which is the only part of that file this repository owns. Its sections are hash-verified against the fleet canonical, so a future agent that adds a rule there in good faith breaks the audit. The warning says so and points at where a rule actually goes. OPERATIONS.md gains a Repository Tooling Hazards section: - A GitHub comment body must never be built inside a double-quoted shell string. Backticks are command substitution, so a body mentioning a path in code formatting executes it. That happened here and ran configure.sh. - configure.sh writes unless given the check argument. - The reference-link convention, the MD033 details/summary allowlist and the nested code element it does not cover, and the widened spell-check scope. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
Pull request overview
Documents two pieces of durable operational knowledge in-repo so they survive across sessions/machines: (1) the byte-locked nature of most AGENTS.md ## sections, and (2) tooling hazards around gh, repo-config/configure.sh, and doc-lint/spell-check conventions.
Changes:
- Add an
AGENTS.mdintro warning that most##sections are carried, hash-verified template content and should not be edited locally. - Add an
OPERATIONS.md“Repository Tooling Hazards” section capturing sharp edges (shell quoting with backticks,configure.shdefault write behavior, and lint/spell-check expectations). - Update the
OPERATIONS.mdopener to direct readers to the new hazards section and clarify where repo-specific durable rules should live.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
OPERATIONS.md |
Adds a “Repository Tooling Hazards” section and updates the intro to point readers at it. |
AGENTS.md |
Adds an intro warning about byte-locked carried ## sections and where repo-specific rules belong. |
The opening sentence read as a blanket prohibition on editing any section, then exempted two of them a clause later. A warning that contradicts itself in the same paragraph is worse than none, and this one risked leaving Repository Layout stale, since an agent told not to edit sections would not update the tree description when the tree moved. Name the exception in the sentence that carries the prohibition, and say plainly that the two intent sections are the repository's to maintain. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The intro claimed this file held everything specific to this repository, and that a durable repo rule goes here. Both overstate it. CODESTYLE.md owns code and documentation style, including the C++ section and the widened spell-check scope, and WORKFLOW.md owns the CI/CD contract, so a style rule filed here would be in the wrong doc and hard to find. Say instead that a repo-specific rule goes to whichever local doc owns its subject, and name the three, keeping only the point that matters: AGENTS.md is byte-locked and cannot take one. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This was referenced Aug 29, 2026
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
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.
Persists the durable knowledge from the hub convergence into the tree, so it survives a new session or a new machine.
Why
The
Durable Knowledge and Self-Improvementsection this repo now carries is explicit that anything a future agent must honor belongs in a committed governance file, not in agent memory. Two things from the convergence qualified and were still only in a chat log.AGENTS.mdintroIts
##sections are hash-verified against the fleet canonical, so a future agent that adds a rule there in good faith fails the audit - and the failure is not obvious from reading the file. The warning goes in the intro, which is the only part ofAGENTS.mdthis repository owns, and points at where a repo-specific rule actually belongs.Verified the edit does not disturb the hashes: all 16 verbatim sections still byte-match, since the audit extracts from
## <heading>onward and the intro sits outside every section.OPERATIONS.md- Repository Tooling Hazardsrepo-config/configure.shwrites unless givencheck. The bare form isapply, which PATCHes settings and PUTs both rulesets.MD033allowlist coveringdetailsandsummarybut not a<code>nested in a<summary>, and the spell-check scope being wider than the fleet default.Verification
Full lint gate clean. All 16 verbatim sections byte-match hub
ea2dd1f, both carried configs match, no undeclared sections.