The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes external content such as git history, project notes, and cycle states, creating a surface where malicious instructions could be embedded in data to influence agent behavior. 1. Ingestion points: Reads live cycle state, plan gates, priority notes, git logs, and file changes as defined in the SKILL.md workflow. 2. Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious content within the ingested data. 3. Capability inventory: The skill specifies it is read-only and only recommends actions without executing them, which mitigates the risk of command execution or data exfiltration. 4. Sanitization: No sanitization or filtering of the read project content is described.
[NO_CODE]: The skill consists entirely of instructional markdown and metadata. It does not include any Python or Node.js scripts, nor does it define external dependencies or shell commands.

nba