The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external build summary files to pre-populate lesson entries, which could allow malicious instructions in those files to influence the agent's behavior or poison persistent project documentation.
Ingestion points: The skill reads SUMMARY-*.md files located in the .shipyard/phases/{N}/results/ directory as described in the 'Pre-Population' section of SKILL.md.
Boundary markers: There are no explicit boundary markers or instructions to treat the ingested data as untrusted text rather than instructions.
Capability inventory: The skill has the capability to read project files and write to persistent documentation files including .shipyard/LESSONS.md and CLAUDE.md. It does not have network access or command execution capabilities.
Sanitization: The skill lacks sanitization, escaping, or filtering of the content extracted from the build summaries before it is presented to the user or written to the filesystem.

lessons-learned