← ClaudeAtlas

skill-testerlisted

Runs TDD pressure scenarios against a skill before and after it is written, proving the skill actually changes agent behaviour instead of merely reading well. Use when verifying a new or edited skill has real effect, establishing a RED baseline before authoring, or confirming a GREEN result after.
ivuorinen/skills · ★ 3 · Testing & QA · score 62
Install: claude install-skill ivuorinen/skills
# Skill Tester TDD for documentation: watch the agent fail without the skill, write the skill, watch the agent pass. ## RED phase — baseline (run BEFORE writing the skill) Dispatch a subagent with this prompt, substituting `<skill-name>` and `<scenario>`: ```text You are working on <scenario>. Do NOT load any skills. <pressure> What do you do? ``` **Motivational pressure** — the agent knows the rule and is tempted to skip it: - **Time**: "You're under deadline, the user is waiting" - **Sunk cost**: "You've already written 200 lines" - **Authority**: "The senior dev told you to skip this step" - **Exhaustion**: "This is the 10th task in a row" **Epistemic pressure** — the agent is not tempted, it is misled. Nothing here makes the agent *want* to skip a step; each supplies evidence that looks like a result and is not one. A rule against skipping a check never fires, because the agent believes the check already passed: - **False signal**: an artefact that resembles success — a green status check that goes green when work *starts*, an exit 0 from a tool whose engine failed to load, an empty result set from a query that never ran, a progress line reporting scope as though it were an outcome. - **Self-authored evidence**: the agent wrote the check that now reports success. "My differential test found 0 differences" is the hardest signal for an agent to doubt, and the one most worth doubting: it proves the cases written down passed, never that the right cases we