The Agent Skills Directory

PROMPT_INJECTION (LOW): Indirect prompt injection surface detected.
Ingestion points: Data enters the agent context from files within .agents/learnings/, .agents/patterns/, .agents/research/, and .agents/retros/ (referenced in SKILL.md).
Boundary markers: Absent. The skill summarizes and injects content directly into the session context without visible delimiters or instructions to ignore embedded commands.
Capability inventory: The skill utilizes the ao inject CLI tool and standard shell commands (ls, head) to retrieve and process local artifacts.
Sanitization: Absent. There is no evidence of escaping or filtering content retrieved from the knowledge pools before it is presented to the agent.
COMMAND_EXECUTION (LOW): The skill executes local CLI tools and shell scripts for knowledge retrieval and validation.
Evidence: SKILL.md documents execution of ao inject --apply-decay and ls -lt for context retrieval.
Evidence: scripts/validate.sh uses bash -c to run validation checks on the skill's structure, though the commands are statically defined.

inject