You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
docs(invocation-mode): re-grade harness-ops:plugins under the 2026-10-18 decision #9708
Part of #6458. Re-grade harness-ops:plugins under the owner's 2026-10-18 decision on #6458 (D3, option b): a destructive-capable skill becomes model-invocable only if every destructive step provably stops for a person even when it runs unattended. The four class (i) skills are graded one at a time.
Scope
Read plugins/harness-ops/skills/plugins/SKILL.md, its context/ pages and scripts. For each destructive or mutating step, quote the line that stops it for a person, and say whether that stop holds with no person present (for example a script flag the model can pass, such as --yes, does not hold).
Update its row in docs/conventions/invocation-mode/README.md (the 2026-08-17 grade table) with the evidence and the grade: FLIP to false only when every step holds, otherwise KEEP true citing the step that does not.
On FLIP, set disable-model-invocation: false (or remove the key) in that SKILL.md and add a changelog fragment for the plugin.
Done when
The harness-ops:plugins row in docs/conventions/invocation-mode/README.md cites the 2026-10-18 decision and quotes each destructive step's stop.
git show HEAD:plugins/harness-ops/skills/plugins/SKILL.md | grep -c '^disable-model-invocation: true' prints 1 for KEEP or 0 for FLIP, matching the row.
Proposed work class: work-class: scoped
Basis: automated intake triage; the lanes App applied this class.
Size: Small
Complexity: Medium
Uncertainty: Knowable
Flags: none
Understood: re-grade the harness-ops:plugins skill under the 2026-10-18 decision on #6458. That means checking each destructive step for a stop that holds with no person present, then updating its row in the invocation-mode README. If every step holds, it also means flipping the SKILL.md key and adding a changelog fragment. The scope and done-when items are clear. The grade depends on reading the skill and its scripts. No likely duplicates found among the open issues I checked.
The work-resolve-issue run that agent-ready started failed or was cancelled before it reported, or it stopped or the token broker refused it because its workflow files differ from main's. work-unblock removed and added agent-ready back to start a fresh run on main's workflow files. If this run ends the same way, it re-fires the issue again, waiting longer each time.
Activity
github-actions commentedon Oct 24, 2026
Proposed work class: work-class: scoped
Basis: automated intake triage; the lanes App applied this class.
Size: Small
Complexity: Medium
Uncertainty: Knowable
Flags: none
Understood: re-grade the
harness-ops:pluginsskill under the 2026-10-18 decision on #6458. That means checking each destructive step for a stop that holds with no person present, then updating its row in the invocation-mode README. If every step holds, it also means flipping the SKILL.md key and adding a changelog fragment. The scope and done-when items are clear. The grade depends on reading the skill and its scripts. No likely duplicates found among the open issues I checked.Automated triage, applied by the lanes App.
melodic-automation-lanes commentedon Oct 24, 2026
Not admitted to the issue lane yet: blocker #9707 is not closed as completed by a listed human or a merge.
melodic-automation-lanes commentedon Oct 25, 2026
The work-resolve-issue run that
agent-readystarted failed or was cancelled before it reported, or it stopped or the token broker refused it because its workflow files differ from main's. work-unblock removed and addedagent-readyback to start a fresh run on main's workflow files. If this run ends the same way, it re-fires the issue again, waiting longer each time.