Document the new gates; make claim_gate's modes actually differ - #750
Merged
Conversation
publish_guard, claim_gate and the agent-authored-validator check shipped in 0.21.6 with no docs. config.md gains the two keys, verification-discipline.md gains a section each, and features.md lists them with the rest of the loop guards. Two documented behaviours had also gone stale: the masked-command guard now treats a newline as the separator it is, and the A/B harness records gate co-occurrence and has a self-test. Both sections say so, including what the newline rule costs — an honest multi-line verification whose last line isn't the check gets declined, same as the ';' rule already did. Also: claim_gate advertised off/advisory/blocking while the wiring only checked != Off, so the last two were byte-identical and the config surface promised a distinction that didn't exist. Advisory is one-shot; blocking re-enters up to three times, mirroring MAX_REVIEW_REACT. A test now fails if they collapse again. mcp-server.md documents the delegate response's evidence object and that files_changed is scoped to the delegation.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The three mechanisms that shipped in 0.21.6 —
publish_guard,claim_gate, and the agent-authored-validator check — had no documentation. This adds it, and fixes something the writing surfaced.Docs
config.md: the two new keys, with what each mode does.verification-discipline.md: a section each. The publish-state guard's entry explains why there is no override token (the source harness shipped an overrideable version first and measured it leaking), and why it blocks discarding but never modifying (a deliberate departure — the source's setting was one-shot submission, dirge sessions keep working after green).features.md: listed alongside the other loop guards, plus the memory expectations.mcp-server.md: the delegate response'sevidenceobject, and thatfiles_changedis now scoped to the delegation.Two sections had gone stale and now say so: the masked-command guard treats a newline as the separator it is, and the A/B harness records gate co-occurrence. The newline entry states what the rule costs rather than only what it catches.
The fix
claim_gateadvertisedoff | advisory | blockingwhile the wiring only checked!= Off— the last two were byte-identical, so the config surface promised a distinction that did not exist. Advisory is one-shot; blocking re-enters up to three times, mirroringMAX_REVIEW_REACT. A test fails if they collapse again.Found by trying to write the docs, which is the usual way.