What problem does it solve? Adding a new watch type to a polling triage loop is error-prone: a checker that mismanages its watermark either silently drops activity or burns money on false dispatches, and nothing catches the mistake. This Skill encodes the full doctrine for writing, registering, testing, and documenting a checker so neither failure mode ships. ## Core Features & Use Cases - Checker authoring doctrine: Defines the result contract (clean, dirty, hold, ratelimited, error, misconfig), watermark advancement rules, credential isolation, and argument-injection defenses for executable checkers/<type>.py scripts. - Registration and manifest guidance: Explains the <type>.watch.json manifest fields, derived consumers, Slack slack_* naming rule, and optional .lag files for eventually consistent sources. - Test and doc coverage checklists: Specifies the minimum test cases (query shape, tie safety, monotonic advance, failure classification) and the documentation tables that must stay in sync, enforced by contract tests. - Use Case: When asked to make the triage loop watch a new source like an issue tracker or queue, follow this Skill to copy the closest sibling checker, write the manifest, prove watermark safety with faked binaries, and update all docs in one commit. ## Quick Start Ask the agent to add a new watch type for your source by following the yaas-checker-authoring skill, starting from the closest existing checker such as github_pr.py.
