Describe rook 0.1.11 in the agent skill

[lt-user-access-token-stage]

Regenerated the Rook agent skill for the 0.1.11 release.

Release	0.1.11 (npm @testmuai/rook@0.1.11, release v0.1.11)
Model	claude-opus-5-5
Files	skill-installer/skills/references/scenarios.md skill-installer/skills/references/verdicts.md skill-installer/skills/references/headless-contract.md

Stated rationale: Documented the release's sync feature check, the registered-session refusal, full session transcripts, OWASP and MITRE scenario references with named unrunnable scenarios, and the report changes for refusals, coverage gaps and partial compromises, while leaving cosmetic, header and text-only changes undocumented.

Behavior accounting (model claims; verify against the diff)

• sync-orphans: edit — Added a note to scenarios.md that sync stops before uploading when a scenario names a feature the agent lacks, lists the fix, and must succeed before an upstream run. (skill-installer/skills/references/scenarios.md)
• registered-session: edit — Added an errors bullet saying such commands refuse before running or charging when the session is unregistered, and should be reported as no run and no spend. (skill-installer/skills/references/headless-contract.md)
• session-transcript: edit — Extended the home-state paragraph to say transcript.jsonl records raw input and output of every role call, including refused, failed and interrupted calls. (skill-installer/skills/references/headless-contract.md)
• client-headers: no-edit — Request headers and debug client display are internal transport details. The skill neither uses nor describes them for driving or interpreting runs. (no applicable skill path)
• owasp-mapping: edit — Added to scenarios.md that red-team scenarios carry OWASP LLM, ASI and MITRE ATLAS references, and that generate writes and names unrunnable scenarios with a reason instead of dropping them. (skill-installer/skills/references/scenarios.md)
• redteam-reports: edit — Added to verdicts.md that a grounded partial compromise counts as Fail and that reports keep refusals, interrupted runs and coverage gaps for the summary. (skill-installer/skills/references/verdicts.md)
• cli-text: no-edit — SKILL.md and ci.md already teach sync before run, and troubleshooting.md already routes the update remedy through rook update. The guide and help wording changes alter no command, flag or result the skill relies on. (skill-installer/skills/SKILL.md, skill-installer/skills/references/ci.md, skill-installer/skills/references/troubleshooting.md)
• logout-note: no-edit — The skill never runs rook logout. ci.md already explains that LT_USERNAME and LT_ACCESS_KEY sign commands in ahead of any stored sign-in. (skill-installer/skills/references/ci.md)
• consent-logo: no-edit — The logo on the consent page and the one-time OAuth client re-registration are cosmetic. They do not change the login commands or remedies the skill documents. (no applicable skill path)
• boot-banner: no-edit — The TUI boot banner mascot is purely visual. The skill drives the CLI headlessly and does not describe the banner. (no applicable skill path)

Generated from the published CHANGELOG section, GitHub release notes and help output of the released
binary installed from public npm — no private source was read. The mirrors were
regenerated with scripts/sync-skill-mirrors.sh, and the skill mirror
and flag harnesses passed against rook 0.1.11 before this PR opened.
The contract harness runs in this PR's public CI, away from the private
checkout and the token that opened the PR.
Another release PR can open concurrently; compare this branch with current main before merging.

Read the prose before merging. The harnesses check format, mirror parity and
that every documented flag exists; none of them checks whether a sentence is true.

[samyakLambda]

I checked the bot's prose against the published @testmuai/rook@0.1.11 and its CHANGELOG. All six "edit" claims are accurate, and the four "no-edit" calls are reasonable.

I added 4ce9689 to fill the gaps it left:

• Scenarios that can't run: generate keeps them with executable: false and a skip_reason starting coverage gap:, invalid scenario: or invalid red-team scenario:. coverage-gaps.yaml lists attack patterns that got no scenario.
• Red-team hooks: patterns that need fixture seeding, a collect hook or a fresh session (open + close) are skipped until the profile has that hook.
• Run cap: run takes at most two scenarios per attack pattern unless --only is used.
• Report fields: report --json can carry coverage_gaps, redteam_coverage and summary_claims. An older stored Partial reads as Unable to Verify.
• Version notes: two "In 0.1.6" notes become "Since 0.1.6". Both still hold in 0.1.11.

Checks against rook 0.1.11:

• test-skill-mirrors passes.
• test-skill-flags: 47 command lines pass.
• test-skill-contract: 16/16.
• Installer tests: 19/19.
• CI is green.

The branch is 2 commits behind main (#53 and #54), but they touch only the README and the mascot assets, so it merges cleanly.

Needs one approval. After the merge, publish-skill releases @testmuai/rook-skill@0.1.11; npm is still on 0.1.7.