The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local CLI tools including git for repository analysis (logs, diffs, and bisection) and the GitHub CLI (gh) for managing pull requests and verification results.
[EXTERNAL_DOWNLOADS]: Retrieves evidence from external well-known telemetry (Dash0) and issue tracking (Linear) platforms via dedicated MCP servers. These downloads are essential for resolving the context of reported bugs.
[PROMPT_INJECTION]: The skill is designed to process untrusted external data such as telemetry logs, stack traces, and ticket descriptions, which introduces a surface for indirect prompt injection. \n- Ingestion points: Phase 0 input classification and Phase 1 evidence resolution procedures. \n- Boundary markers: Utilizes a structured 'Evidence Record' and a persistent 'Bug-notes ledger' to isolate and track data across execution phases. \n- Capability inventory: Dispatches implementation tasks to sub-agents (aw-executor) capable of workspace modifications and test execution. \n- Sanitization: Mitigates risks through a multi-stage pipeline, including independent verification by a 'bug-fix-verifier' agent in a fresh context with no access to prior reasoning.

fix-bug