agentleFS
joris887joris887/exosuit/.claude/skills/claude-sense-check/SKILL.md

claude-sense-check

Skill9 starsChanged 3 months agoMITView on GitHub

Incremental code logic verification of UAT test cases — pick batch, trace code, fix issues, update tracking.

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. About 0 are common, 14 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈14

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    If all cases are checked, report "All UAT cases have been sense-checked" and stop

    line 24
  2. Distinctive:

    Before checking any logic, collect the full picture for the batch's feature area:

    line 40
  3. Distinctive:

    Read all source files relevant to the batch

    line 42
  4. Distinctive:

    Read existing tests for the relevant code — they reveal intent and edge cases

    line 45
  5. Distinctive:

    Check the Covers field on each test case to identify which backlog stories are involved

    line 46
  6. Distinctive:

    Read the story acceptance criteria from the relevant epic/backlog file

    line 47
  7. Distinctive:

    Do NOT proceed to logic checks until you have the full picture.

    line 49
  8. Distinctive:

    Read the test case steps and acceptance criteria carefully

    line 55
  9. Very specific:

    Check the Claude Sense Check checkbox: - [x] Logic verified from code perspective

    line 78
  10. Distinctive:

    After updating the tracking file, update any status summary tables to reflect the new sense-check counts.

    line 86
  11. Distinctive:

    Never re-check an already checked case

    line 129
  12. Distinctive:

    Never check more than 5 cases in one run — stay focused

    line 130
  13. Distinctive:

    Always update the tracking file before finishing — the checkboxes are the source of truth

    line 131
  14. Distinctive:

    Always gather full context before starting logic checks — no guessing

    line 132
  15. Very specific:

    Always use /story-cycle for fixes — maintain TDD discipline

    line 133
  16. Very specific:

    Follow coding standards in docs/reference/CODING_STANDARDS.md

    line 134
  17. Distinctive:

    Follow verification rules: run the project's test command before claiming fixes are complete

    line 135

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

claude-sense-check

Performing incremental Claude Sense Check on UAT backlog.

Phase 1: Read Current State

  1. Find the UAT tracking file: check for docs/testing/UAT_COVERAGE.md, docs/testing/UAT_INDEX.md, or docs/testing/uat/ directory
  2. Scan for Claude Sense Check sections in every test case
  3. Build a work queue: cases where - [ ] Logic verified from code perspective is unchecked
  4. Skip cases where the checkbox is already [x]
  5. If all cases are checked, report "All UAT cases have been sense-checked" and stop

Phase 2: Pick Next Batch

From the unchecked cases, sorted by priority (critical > high > medium > low), then by ID within each priority level:

  1. Select 2-5 cases that share the same module, feature area, or user flow
  2. Present to the user:
    • Batch: which cases (e.g., UAT-005, UAT-006, UAT-007)
    • Area: the functional area they belong to
    • Rationale: why they belong together
    • Relevant files: source files and components expected to be involved
  3. Proceed after presenting (no user confirmation needed — this is a code-only check)

Phase 3: Gather Context

Before checking any logic, collect the full picture for the batch's feature area:

  1. Read all source files relevant to the batch
  2. Trace the data flow for the feature area end-to-end
  3. Note dependencies, shared state, API contracts, and external integrations
  4. Read existing tests for the relevant code — they reveal intent and edge cases
  5. Check the Covers field on each test case to identify which backlog stories are involved
  6. Read the story acceptance criteria from the relevant epic/backlog file

Do NOT proceed to logic checks until you have the full picture.

Phase 4: Code Logic Check (per case)

For each case in the batch:

  1. Read the test case steps and acceptance criteria carefully
  2. Trace each step through the actual source code
  3. For each acceptance criterion, verify:
    • The code path exists and is reachable
    • Conditionals are correct (no off-by-one, no inverted logic)
    • Data flows correctly between components
    • Error handling covers the expected failure modes
    • Edge cases mentioned in the test case are handled
  4. Assign a verdict with confidence (high = full code path traced, medium = partial trace or complex logic, low = inference without full trace):
    • Pass (high/medium/low) — logic is correct, code handles the scenario properly
    • Warning (high/medium/low) — potential issue found, describe with file:line reference
    • Fail (high/medium/low) — confirmed broken, describe with file:line reference and root cause

Phase 5: Resolve Issues

For every Warning or Fail verdict, classify and act. Read references/issue-classification.md for the full classification guide (quick fix vs design gap).

Phase 6: Update Tracking

Per-case: Update the UAT tracking file

For each case in the batch:

  1. Check the Claude Sense Check checkbox: - [x] Logic verified from code perspective
  2. Fill in the Notes line with verdict and what was checked or fixed:
    • Pass: "Pass (high) — verified [what was checked], code handles [scenario] correctly"
    • Fixed: "Fixed (high) — [issue description], resolved in [commit-hash]"
    • Design gap: "Warning (medium) -> [story-id] — [gap description], backlog story created"

Per-area: Update UAT index

After updating the tracking file, update any status summary tables to reflect the new sense-check counts.

Phase 7: Report & Hand Off

Output a summary:

### Claude Sense Check Complete

**Batch:** UAT-### to UAT-### ([area name])
**Results:**
| Case | Title | Verdict | Confidence | Notes |
| ---- | ----- | ------- | ---------- | ----- |
| UAT-### | [title] | Pass/Warning/Fixed | high/medium/low | [one-line summary] |

**Fixes Applied:** [count]
- [commit-hash]: [one-line description] (for UAT-###)

**Backlog Stories Created:** [count]
- [story-id]: [one-line description] (for UAT-###)

**Progress:** [x]/[total] UAT cases sense-checked
**Remaining:** [count] cases unchecked
**Suggested Next Batch:** UAT-### to UAT-### ([area name]) — [why]

Include a YAML block for machine-parseable results:

sense_check:
  batch: [UAT-###, UAT-###]
  area: "[area name]"
  results:
    - id: UAT-###
      verdict: pass
      confidence: high
      notes: "[one-line]"
  fixes: 0
  stories_created: 0
  progress: "X/Y"

Rules

  • Never re-check an already checked case
  • Never check more than 5 cases in one run — stay focused
  • Always update the tracking file before finishing — the checkboxes are the source of truth
  • Always gather full context before starting logic checks — no guessing
  • Always use /story-cycle for fixes — maintain TDD discipline
  • Follow coding standards in docs/reference/CODING_STANDARDS.md
  • Follow verification rules: run the project's test command before claiming fixes are complete

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.