Bug Test Skill

Test Bug Fix

Validate that the fix recorded by /speckit-bug-fix actually resolves the bug described by /speckit-bug-assess. The output is a verification report at .specify/bugs/<slug>/test.md.

User Input

$ARGUMENTS

The user input should identify the bug to validate. Accept any of:

• slug=<bug-slug> or --slug <bug-slug> or a bare slug-like token.
• A path that contains the slug (e.g. .specify/bugs/login-timeout/).
• Nothing — fall back to context (see below).

Slug Resolution

Resolve BUG_SLUG in this order, stopping at the first match:

1. Explicit user input — a slug passed in $ARGUMENTS (any of the forms above).
2. Conversation context — if the current session has just run /speckit-bug-assess or /speckit-bug-fix, the slug it reported is the working slug. Reuse it without re-prompting. Confirm it by checking that .specify/bugs/<slug>/fix.md exists; if it does not, fall through.
3. Single candidate on disk — list .specify/bugs/*/fix.md. If exactly one bug has a fix.md, use it.
4. Disambiguate:
   • Interactive mode: ask the user which bug to validate and list the candidates.
   • Automated mode: stop with an error listing the candidates. Do not guess.

Once resolved, set BUG_SLUG and BUG_DIR = .specify/bugs/<BUG_SLUG>, and briefly state in your reply which resolution path was used (explicit / from context / single candidate / asked).

Prerequisites

• BUG_DIR/assessment.md MUST exist.
• BUG_DIR/fix.md MUST exist. If not, stop and instruct the user to run /speckit-bug-fix first.
• If BUG_DIR/test.md already exists, ask the user whether to overwrite it (interactive mode) or refuse (automated mode).
• Read both assessment.md and fix.md in full so you know:
  • The original symptom and reproduction steps (from assessment.md).
  • The actual code changes and tests added (from fix.md).

Execution

1. Plan the validation

   • Decide which checks prove the bug is gone:
     • Re-run the reproduction steps from the assessment (or their automated equivalent).
     • Run the tests added or updated in the fix.
     • Run any broader regression suite that touches the changed files.
   • Decide which checks prove nothing was broken:
     • Existing test suites for the changed modules.
     • Lint / type-check if the project uses them.

2. Run the checks

   • Execute each planned check. Capture command, exit status, and a short excerpt of relevant output (last few lines, or the failing assertion).
   • If a check is destructive, network-dependent, or expensive, skip it and record it as skipped with a reason; do not run it without explicit user consent.
   • If you cannot run a check at all (missing tooling, no test framework configured), record it as not-run with a reason instead of fabricating a result.

3. Judge the outcome

   • Mark the fix as:
     • verified — all critical checks pass and the original symptom no longer reproduces.
     • partial — the original symptom is gone but unrelated regressions appeared, or some checks are inconclusive.
     • failed — the symptom still reproduces or the regression suite is broken by the fix.
   • Do not over-claim. If reproduction was not actually performed (e.g., the bug required a production environment), say so explicitly.

4. Write the verification report

   Write to BUG_DIR/test.md using this structure:

   # Bug Verification: <short title>
   
   - **Slug**: <BUG_SLUG>
   - **Tested**: <ISO 8601 date>
   - **Assessment**: ./assessment.md
   - **Fix**: ./fix.md
   - **Result**: verified | partial | failed
   
   ## Summary
   
   <One or two sentences: does the bug reproduce, did the fix hold, were any regressions found.>
   
   ## Checks Performed
   
   | Check | Command / Action | Result | Notes |
   |-------|------------------|--------|-------|
   | Reproduction (post-fix) | <command or manual steps> | pass / fail / skipped / not-run | <short note> |
   | New / updated tests | `<command>` | pass / fail | <short note> |
   | Regression suite | `<command>` | pass / fail / skipped | <short note> |
   | Lint / type-check | `<command>` | pass / fail / skipped | <short note> |
   
   ## Output Excerpts
   
   <Short snippets of relevant output (e.g., final summary line of a test run, the failing assertion). Keep it tight — no full logs.>
   
   ## Residual Risks
   
   - <known limitation, environment not covered, etc.>
   
   ## Recommendation
   
   <One paragraph. Examples:>
   - "Close the bug — verified end-to-end."
   - "Hold — reproduction inconclusive; needs verification in staging."
   - "Reopen — symptom still reproduces; rerun `/speckit-bug-assess`."
   

5. Report back with:

   • The slug and BUG_DIR/test.md path.
   • The result (verified, partial, failed).
   • If the result is failed, recommend re-running /speckit-bug-assess with the new evidence captured in test.md.

Guardrails

• This command MUST NOT modify source code. It only runs checks and writes inside .specify/bugs/<slug>/.
• Never overwrite an existing test.md without confirmation.
• Never mark a fix as verified based on tests alone if the original assessment listed a reproduction that you did not actually exercise — downgrade to partial and say so.

Done When

• The original reproduction is re-run and now passes, quoted in the report
• The regression test fails when the fix is reverted — verified, not assumed
• Adjacent behavior checked for collateral damage