The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from various local repositories (~/repos/*), PRDs, and user conversations to define a global mission. If these source files contain malicious instructions, they could influence the generated mission and be propagated to other projects.
Ingestion points: The skill reads ~/repos/*/CHOICES.md, README.md, and external objective/metric documents during the detection and interview phases.
Boundary markers: There are no explicit markers or 'ignore embedded instructions' directives provided when processing the repository content or external PRDs.
Capability inventory: The skill possesses the ability to write to ~/vault/missions.md and dispatches the /apply-mission skill, which likely performs file writes across multiple repositories.
Sanitization: The skill does not specify methods for sanitizing, escaping, or validating the content read from local repositories before incorporating it into the global mission definition or the missions.md file.

define-mission