The Agent Skills Directory

[PROMPT_INJECTION]: Indirect prompt injection surface identified. The skill ingests untrusted data from the target project which could be used to influence the agent's behavior.
Ingestion points: Reads external data from README.md (via head), git log output, and manifest files like package.json (via jq).
Boundary markers: Results are formatted into a structured summary template, but the skill does not utilize specific delimiters or defensive instructions to prevent the agent from obeying commands embedded within the analyzed files.
Capability inventory: The skill is authorized to use Bash, Read, Grep, and TodoWrite tools.
Sanitization: No explicit sanitization or filtering of instructions within the ingested content is performed beyond standard shell utility output processing.
[COMMAND_EXECUTION]: Uses local shell scripts (discover.sh and analyze-skills.sh) to execute environment-gathering commands including git, jq, ls, and find. These commands are used strictly for project metadata extraction and orientation within the local file system.

project-discovery