The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub issues, including titles, bodies, and comments, which are used to drive the orchestration logic. Maliciously crafted issue content could potentially influence the agent's behavior.
Ingestion points: GitHub issue metadata and content processed in modules/intake.md.
Boundary markers: The instructions do not specify the use of delimiters or protective prompt wrappers when handling external data.
Capability inventory: Extensive capabilities including shell command execution (git, gh, python3), spawning subagents, and invoking other specialized skills as seen in SKILL.md.
Sanitization: No explicit sanitization or content validation is mentioned beyond checking for requirement specificity.
[COMMAND_EXECUTION]: The orchestrator performs various shell operations such as git checkout, gh issue view, and executes local Python scripts (night_run.py, deferred_capture.py). These commands are integral to the skill's function but represent an attack surface if manipulated.
[PERSISTENCE]: The skill implements a self-healing heartbeat using CronCreate with the recurring: true parameter to monitor and resume the orchestration loop every 5 minutes.
[EXTERNAL_DOWNLOADS]: The skill communicates with GitHub APIs via the gh CLI tool to retrieve issue data. This is documented behavior for the intake process as described in modules/intake.md.

summon