| name | finstack-quality-gate-triage |
| description | Diagnoses and fixes pasted finstack quality-gate failures, including mise pre-commit-run, Rust clippy/test failures, cargo deny, OSV, Python build/lint/type/parity failures, WASM build/test failures, and iterative bug-hunting loops. Use when the user shares terminal output, asks to fix CI/pre-commit/lint/test failures, or asks to find and fix bugs until checks are clean. |
Quality Gate Triage
Use this skill when the deliverable is a clean failing check or an iterative bug-fix loop. Prefer narrow evidence from the failing command over broad speculation.
Routing
Use this skill for:
- pasted compiler, clippy, lint, typecheck, test, parity,
cargo deny, OSV, maturin, WASM, or pre-commit output,
- requests like "fix this failure", "pre-commit is red", "CI failed", "run a bug hunt", or "keep going until the check is clean",
- bugs discovered while implementing another plan when the next step is blocked on a quality gate.
Use finstack-quant-finance-review first when the failure points to pricing/risk correctness. Use finstack-binding-parity-reviewer first when the failure is primarily cross-language API drift.
Workflow
- Capture the failing command and the first real error.
- Identify the affected component: Rust, Python, WASM, bindings, docs, dependency audit, or parity.
- Read the smallest relevant files and tests.
- Fix the root cause, not downstream symptoms.
- Re-run the smallest check that proves the fix.
- If a new failure appears, repeat from the new first real error.
- Stop only when the target check passes or a blocker requires user input.
Finstack Command Map
Prefer repo tasks from AGENTS.md and mise.toml:
- Rust:
mise run rust-lint, mise run rust-test
- Python:
mise run python-build, mise run python-lint, mise run python-test, mise run python-typecheck
- WASM:
mise run wasm-build, mise run wasm-lint, mise run wasm-test
- Broad gates:
mise run all-lint, mise run all-test, mise run all-ci
- Pre-commit:
mise pre-commit-run when the user is already running that gate
Use targeted checks while iterating. Reserve broad checks for broad changes or final verification.
Bug-Hunting Mode
When the user asks for a bug hunt rather than a specific failing check:
- Pick a focused scope instead of scanning the whole workspace blindly.
- Run available static checks for that scope.
- Trace suspicious code through callers and tests.
- Fix real bugs in small slices.
- Re-read touched files with fresh eyes.
- Verify with targeted tests and report residual risk.
Output Format
## Triage Result
### Root cause
<first real error and why it happened>
### Fix
<files changed and behavior restored>
### Verification
- `<command>`: pass/fail
### Residual risk
<unchecked broad gates or unrelated blockers>
Resources
references/TRIAGE.md - triage decision tree.references/PATTERNS.md - common bug patterns.references/TOOLS.md - tool-specific notes.outputs/quality-gate-report.md - example completed triage report.
