The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution to set up environments, run tests, and trace code history.
Evidence: SKILL.md instructs the agent to run commands such as npm install, bun install, bundle install, git log, and git bisect during the investigation and fix phases.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection due to its integration with external issue trackers.
Ingestion points: The agent fetches issue titles, bodies, and full comment threads from GitHub, Linear, and Jira (Phase 0).
Boundary markers: While the initial user input is wrapped in <bug_description> tags, content fetched from remote issue trackers is read into the context without clear delimiters or 'ignore' instructions.
Capability inventory: The skill has broad capabilities including file modification (Phase 3), shell command execution (Phase 1, 3), and opening Pull Requests (Phase 4).
Sanitization: There is no evidence of sanitization or filtering of the external data fetched from issue trackers before it is processed by the agent.
[EXTERNAL_DOWNLOADS]: The skill connects to remote services to fetch context and manage project dependencies.
Evidence: It uses the gh CLI and other web tools to retrieve data from GitHub, Jira, and Linear. It also triggers dependency resolution and downloads via package managers like npm and bun.

ce-debug