/flake-watch¶
Testing class
Writes test source into the repo and executes it — the only class that does both. Shared rules: CONVENTIONS-testing.md.
Use this skill when the user asks \"which tests are flaky\", \"what should we quarantine\", \"show the flake report\", or wants per-spec failure history rather than a single run's result. Reads the committed flake ledger and reports fail rate per spec over the last N runs, which specs are trending worse, and which cross the quarantine threshold. Quarantine is always a PROPOSAL carrying an SLA and a tracking issue — never applied silently. Advises only: it writes no test and skips nothing.
At a glance¶
| Run it | /flake-watch |
| Class | testing |
| Version | 0.1.0 |
| Author | navjyotnishant |
| Needs | the flake ledger (§T14) · git · a connected PM tracker (MCP) |
| Source | skills/flake-watch/SKILL.md |
What it needs, and what happens without it¶
Every tool is detected at runtime — none is installed for you.
| Tool | Without it |
|---|---|
the flake ledger (§T14) |
SKIP with a labeled reason — there is no history to report, and inventing one is worse than reporting none |
git |
report rates without the timeline, and say so |
| a connected PM tracker (MCP) | print the issue as paste-ready markdown (§A5 detect-never-require) |
Agents it spawns¶
This skill spawns no subagents — it runs inline.
The procedure¶
The executable steps live in the skill file itself and are deliberately not reproduced here: they are instructions to the model at runtime, not documentation.