The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources such as Dash0 telemetry URLs and Linear ticket descriptions, which could serve as vectors for indirect prompt injection. * Ingestion points: Phase 0 (Intake) and Phase 1 (Evidence Resolution) in SKILL.md and rules/evidence-resolution.md. * Boundary markers: The skill uses a structured Evidence Record and Bug-notes ledger to pass data between phases, providing a degree of isolation from raw inputs. * Capability inventory: The system can perform file system modifications, execute tests, and interact with the GitHub API via the gh CLI. * Sanitization: Key mitigations include delegating root-cause analysis to an isolated sub-agent (rca-investigator) and utilizing an independent verifier (bug-fix-verifier) in a fresh context.
[COMMAND_EXECUTION]: The workflow relies on standard development tools including the GitHub CLI and git for branch management, worktree creation, and pull request operations. * Evidence: Usage of gh pr view, gh pr ready, git worktree, and git bisect is documented in rules/autonomous-handoff.md and rules/reproduction.md.
[EXTERNAL_DOWNLOADS]: The skill retrieves telemetry data and ticket information from external services Dash0 and Linear via dedicated MCP servers. * Evidence: Procedural logic for fetching evidence from these services is detailed in rules/evidence-resolution.md.

fix-bug