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
Course lane 8's fleet grade (docs/conventions/invocation-mode/README.md § Fleet grade, 2026-08-17) examined the non-setup disable-model-invocation: true skills against the rubric's three exception classes. session-flow:show-options landed 2026-08-18 (b3e7af72), a day after that grade, so it was never in the graded population — the ADR 0005 bound scopes the grade to the question asked at that date, not to skills added later.
Surfaced during #2969 (PR #3023) by an automated reviewer: a post-flip recount showed 10 non-setup true skills against a table carrying 9 verdicts. The gap is now recorded in the rubric doc rather than left implicit, and this issue carries the grading.
skill-quality:check check 24 (shipped in #2968) does not fail this — a non-setup true gets a hand-verify note, precisely because class attribution is not machine-checkable.
Proposed work
Grade session-flow:show-options against the rubric's three exception classes: (i) side-effect / manual-timing workflow, (ii) setup skill, (iii) maintainer-only.
Its stated job is laying out fitting skills as a decidable, nothing-hidden menu for the human to choose from. Note the rubric's router-skill verdict (REJECTED, 2026-08-17): a model-invoked skill that routes the agent to user-invoked skills is rejected for this fleet, with a carve-out for domain-scoped composition routers. Whether show-options is that rejected pattern, the admitted carve-out, or a class-(i) deliberate-mode-entry is exactly the question to settle — do not assume the answer from its current true.
Context
Course lane 8's fleet grade (
docs/conventions/invocation-mode/README.md§ Fleet grade, 2026-08-17) examined the non-setupdisable-model-invocation: trueskills against the rubric's three exception classes.session-flow:show-optionslanded 2026-08-18 (b3e7af72), a day after that grade, so it was never in the graded population — the ADR 0005 bound scopes the grade to the question asked at that date, not to skills added later.Surfaced during #2969 (PR #3023) by an automated reviewer: a post-flip recount showed 10 non-setup
trueskills against a table carrying 9 verdicts. The gap is now recorded in the rubric doc rather than left implicit, and this issue carries the grading.skill-quality:checkcheck 24 (shipped in #2968) does not fail this — a non-setuptruegets a hand-verify note, precisely because class attribution is not machine-checkable.Proposed work
session-flow:show-optionsagainst the rubric's three exception classes: (i) side-effect / manual-timing workflow, (ii) setup skill, (iii) maintainer-only.show-optionsis that rejected pattern, the admitted carve-out, or a class-(i) deliberate-mode-entry is exactly the question to settle — do not assume the answer from its currenttrue.falseif no class fits (the planning: flip questionnaire to model-invoked — the one invocation-mode re-grade flip #2969 precedent).Acceptance criteria
session-flow:show-optionscarries a recorded verdict indocs/conventions/invocation-mode/README.md(a class + KEEPtrue, or a flip tofalse)References
docs/conventions/invocation-mode/README.md(the rubric and its fleet grade)plugins/session-flow/skills/show-options/SKILL.md