The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it instructs the agent to read and act upon content from local project files.
Ingestion points: The agent reads state and context from UNITS.csv, STATUS.md, DECISIONS.md, and files in the pipelines/ directory.
Boundary markers: The instructions lack explicit delimiters or warnings to ignore potentially malicious instructions embedded within these project files.
Capability inventory: The skill involves reading multiple project files, writing updates to DECISIONS.md, and potentially invoking the pipeline-router command as indicated in the troubleshooting section.
Sanitization: No content validation or sanitization is specified for the inputs derived from the external files.
[NO_CODE]: This skill consists entirely of natural language instructions in Markdown format and does not include any executable scripts, binaries, or configuration files that run code directly.

human-checkpoint