Skill-invocation guard on plan/research file edits should not fire once the plan already has a finalized status and decision

[DrumRobot]

Problem

A PostToolUse guard fires automatically whenever a plan or research markdown file is edited, and instructs the assistant to invoke a specific skill afterward. The guard's purpose is to make sure in-progress plan edits go through that skill's review flow.

However, the guard fires unconditionally on every edit, including edits made after a plan has already reached a finalized state — for example, a plan file whose frontmatter already records status: ready and a recorded decision: value, being edited only to relocate it to a different repository or to correct a file path reference. In a single session this guard fired 6+ times for edits of that kind, none of which needed the review flow the guard exists to protect. Each time, the edit's intent was still validated (informally, by re-checking the change made sense) but the guard's own literal instruction to invoke the skill was not followed, since doing so for a purely mechanical fix to an already-decided plan adds no value.

Scope

1. Add an exemption condition to the trigger source (the part of the codebase that generates the guard) so that when a plan/research file's frontmatter already declares both a finalized status (e.g. ready) and a recorded decision field, edits to that file do not re-trigger the skill-invocation guard.
2. Regenerate the compiled guard from the updated trigger source using the project's existing trigger-compilation tooling.
3. The trigger compiler currently detects "content worth guarding" via a hardcoded pattern matched against checklist/checkbox-style diffs only. Extending it to also recognize frontmatter field state (status/decision) requires understanding the compiler's current matching approach first — a plain regex addition on top of the existing checkbox-diff pattern is not sufficient by itself.

Verification

Feature	Procedure	Expected Result
Guard exempts finalized plans	Edit a plan file whose frontmatter already has status: ready and a decision: field (e.g. a relocation-only or path-correction edit)	Guard does not fire the skill-invocation instruction
Guard still fires for undecided plans	Edit a plan file that does NOT yet have both status: ready and a decision: field	Guard fires as before, unchanged
No regression in existing checkbox-diff detection	Edit a plan file's checklist/checkbox content	Guard still fires per its original behavior