pedrohcgs/claude-code-my-workflow / .claude/skills/vaccinate
vaccinate
VerifiedQualify a check before it is allowed to clear anything — prove it can detect the failure it is meant to catch. Seeds known defects into a copy of a real artifact plus a clean control, runs the checker, and reports recall and false-positive rate into a qualification ledger. Use when the user says "does this check work", "qualify the gate", "test my reviewer", "seed defects", "vaccinate", "qualify the checks", "can I trust this review", "does it pass for the right reason", or before relying on any automated check or referee simulation for a decision that matters. NOT a code fixer and NOT a reviewer itself — it grades the grader.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.4 KB |
| evals/README.md | markdown | 5.9 KB |
| evals/cases/README.md | markdown | 1.6 KB |
| evals/cases/detects-stale-gate.md | markdown | 442 B |
| evals/cases/nottrigger-simple-question.md | markdown | 186 B |
| evals/cases/requires-clean-control.md | markdown | 277 B |
| references/defect-library.md | markdown | 4.2 KB |
evals/cases/README.md
View on GitHub# Eval cases
One file per case. The prompt is everything between `## Prompt` and `## Assert`; assertions
are the `- ` lines from `## Assert` **to the next heading** — so keep notes either above
`## Prompt`, or under their own heading after the Assert block.
```markdown
## Prompt
The exact user message to send, verbatim. Write what a real user would type,
not a well-specified instruction — the point is to test the skill's triggering
and its instructions, not your prompt-writing.
## Assert
- a literal substring that must appear in a correct answer
- another one
```
**Assertions are literal substrings, matched case-insensitively.** That is crude on
purpose: a grader that is itself a model introduces the failure mode being measured. If a
case cannot be graded by substring, it probably is not a case — it is a judgement, and
judgements belong to you.
## Writing cases that measure something
- **Realistic prompts.** "review this" beats "using the seven-pass protocol, review this",
because the second tests your memory of the skill rather than the skill.
- **Assertions on substance, not shape.** Assert that a specific defect is named, not that
the answer contains "## Findings".
- **Include should-NOT-trigger cases.** A skill that fires on everything is as broken as
one that never fires. Name them `nottrigger-*.md` and assert the skill's signature output
is *absent*.
- **Two or three replicates.** A single run does not separate a real improvement from
sampling noise.
## Priority order
Rank by **cost of being wrong**, not by frequency of use:
`/review-paper --peer` → `claim-verifier` → `/audit-reproducibility` → `/challenge`.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |