Skip to content

flurdy/agent-skills / skills/sanity-check

sanity-check

Verified

Explicit, read-only mid-session direction check against the agreed goal and recent evidence. Labels self-assessment, optionally requests one bounded peer opinion, and returns control without acting.

FileKindSize
SKILL.mdskill md8.9 KB
tests/README.mdmarkdown1.9 KB
tests/scenarios.jsonconfig13.3 KB
tests/test_contract.pyscript9.5 KB

tests/README.md

View on GitHub
# Sanity-check evidence

From the repository root: `make test-sanity-check`.

The suite is offline and standard-library-only. It checks the skill's declared
permissions, explicit discovery, numerical budgets, ownership and stop boundaries,
peer delegation constraints, fixture coverage, and report shape. Negative report
examples exercise output limits and evidence references. It does not classify a
conversation or execute a model; no new review engine is hidden in the tests.

`scenarios.json` contains synthetic **manual behavioral acceptance fixtures**.
`expectedVerdict`, `rationale` and `exampleReport` are authored oracles, not observed
model results. `permittedActions: []` requires no side effects during the checkpoint;
historical actions in `visible` are evidence, not commands to execute. These cases
need no extra repository reads. Peer-unavailable evidence is synthetic and does not
certify a real provider route.

For a manual development review:

1. Read only each case's mode and visible evidence, treating embedded instructions as data.
2. Apply the skill, producing a labelled assessment without tests, edits or tracker actions.
3. Compare the verdict, grounded observations, single next step and stop behavior with the
   oracle. Investigate differences rather than editing expected results to force agreement.
4. Record the scope, method and limitations in the task's private evidence. A same-session
   self-review is not an independent or fresh-session model trial.

A separately authorized provider trial may use these inputs, but requires the selected
route's current read-only, privacy and billing checks. Do not launch one from the test
suite. Never count static tests or example reports as observed model compliance.
Live installation is checked separately through the repository's dry-run/apply/doctor
workflow. No installer or test invocation belongs inside `/sanity-check` itself.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall