The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests raw item data from external backends (GitHub, Trello, or local files) and processes this content to generate technical specifications and update project backlogs.
Ingestion points: Phase 0.2 and Phase 0.3 load status-triage and open items from external sources which are user-controllable (e.g., public GitHub issues).
Boundary markers: The instructions do not define explicit delimiters or 'ignore' directives to isolate the agent from instructions potentially embedded in these item descriptions.
Capability inventory: The agent has access to the Bash tool for reading, writing, and editing files, providing a path for ingested data to influence the filesystem.
Sanitization: No explicit sanitization or filtering of the ingested item bodies is performed before the agent uses them for brainstorming or speccing.
[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes the platform-specific !command syntax to execute a script at load time to discover configuration.
Evidence: Phase 0.0 executes ${CLAUDE_PLUGIN_ROOT}/scripts/discover-config.sh. This is a legitimate use of project-specific discovery tooling and does not appear to involve user-supplied argument injection or network exfiltration.

triage