The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to read and analyze session logs, which serves as a potential entry point for indirect prompt injection. If session logs contain attacker-controlled data, the agent might inadvertently adopt malicious instructions when formulating suggestions for the repository's steering files.
Ingestion points: Reads and processes session logs from the local machine in SKILL.md.
Boundary markers: Absent. The instructions do not specify the use of delimiters or distinct sections to isolate log content from the agent's operational logic.
Capability inventory: The skill suggests modifications to critical repository configuration files like CLAUDE.md, AGENTS.md, and CODING_STANDARDS.md.
Sanitization: Absent. There are no instructions provided to filter, validate, or sanitize the content found within the logs before it is used to generate recommendations.

retro