jaktestowac/awesome-copilot-for-testers / plugins/unslop-answers/skills/unslop-answers
unslop-answers
VerifiedCuts AI tells from what the agent says about its own work: completion claimed without a run, invented file paths and API names, terminal output that was never produced, findings with no location, percentages with no denominator, absence claimed from one grep, effort narration instead of results, hedge stacking, buried blockers, puffery in a bug title. Use whenever reporting a result, a fix, a review finding, a test outcome, a coverage or flake number, a root cause, or a release recommendation, and when the request mentions "prove it", "did you actually run it", "be specific", "no fluff", "stop hedging", or "is this real". Must always apply to answers about your own work.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 15.1 KB |
| resources/answer-audit-checklist.md | markdown | 3.2 KB |
| resources/before-after-answers.md | markdown | 7.1 KB |
| resources/claim-types-and-evidence.md | markdown | 6.4 KB |
| resources/style-tells.md | markdown | 6.7 KB |
resources/answer-audit-checklist.md
View on GitHub# Answer Audit Checklist
Run against the finished draft, before it is sent. Every box is a question about this draft, not a general principle.
## Claims
- [ ] Every sentence a reader would act on is at rung 4 of the ladder, or carries its rung
- [ ] "Fixed" appears only where the failure was seen before the change and the pass after it
- [ ] "Tests pass" states which tests ran, how many, and how many were skipped
- [ ] "Verified" is followed by the command that a reader could repeat
- [ ] Every claim of absence quotes the search behind it
- [ ] A root cause is called a hypothesis unless it reproduces on demand
- [ ] No result is passed on from a tool or a subagent without attribution or a check
- [ ] Every command that failed on the way is mentioned, not only the one that worked
## Specifics
- [ ] Every `file:line` in the draft was opened and resolves
- [ ] Every API, matcher, and method exists in the installed version
- [ ] Every flag, config key, and environment variable was checked against the real config
- [ ] Every code block containing output is a real paste, not a reconstruction
- [ ] Every link, issue number, and version was verified or dropped
## Numbers
- [ ] Every percentage names its denominator, its tool, and its metric
- [ ] Every "improved" or "faster" gives both values
- [ ] Every rate gives the run count and the window
- [ ] Every timing gives the number of runs and the spread, or is called one observation
- [ ] No number in the draft was estimated
## Findings
- [ ] Every finding has a location
- [ ] Every plural has a count, or says it was not counted
- [ ] Every severity has an impact clause: who, what they see, how often, workaround
- [ ] Every recommendation says what to do next
- [ ] "Looks good" is backed by what was read in full, sampled, and skipped
- [ ] Tier 1 findings appear above Tier 2 and Tier 3
## Honesty about scope
- [ ] Everything that was asked for is either done or named as not done
- [ ] The blocker, if there is one, is in the first two lines
- [ ] Extra changes beyond the request are listed separately from the fix
- [ ] Assumptions made on the reader's behalf are stated as assumptions
- [ ] Sampling is described as sampling
## Sources
- [ ] "Best practice", "the docs", "the standard" either link a section or are gone
- [ ] Any convention cited names the failure it prevents
- [ ] No invented consensus about what most teams do
## Calibration
- [ ] Measured facts are stated plainly, with no hedging
- [ ] Unrun claims carry no certainty language
- [ ] The measured and the guessed are visibly different in tone
## Noise
- [ ] No narration of the process
- [ ] No file or step counts standing in for a conclusion
- [ ] No "carefully", "thoroughly", "comprehensively"
- [ ] The opening does not restate the request
- [ ] The closing does not summarise the bullets above it
- [ ] No emoji status markers, no table carrying two facts, no bold label repeating its own line
- [ ] No sycophancy, no apology paragraph
- [ ] The closing states which numbers moved, or there is no closing
## The last three
- [ ] Which sentence would a reader act on, and was it actually verified
- [ ] What was not done, and where does the draft say so
- [ ] What in here was typed from memory rather than read
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: 323k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 323k | 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 |