The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands and a specific orchestration script (~/.claude-octopus/plugin/scripts/orchestrate.sh) to perform project validation and create delivery summaries.
[DATA_EXFILTRATION]: The skill reads internal project state files including STATE.md, PROJECT.md, ROADMAP.md, and ISSUES.md. This content is used for archival and summary generation, representing a data access and exposure surface.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface through its reliance on external project data. Ingestion points: Reads from multiple markdown files in the .octo/ directory and the audit results in ~/.claude-octopus/results/, as well as user-provided feedback for lessons learned. Boundary markers: Absent; there are no delimiters used to differentiate between instructions and data in the ingested files or the audit output. Capability inventory: The skill can execute local scripts, modify the file system using sed and cat, and perform git operations. Sanitization: Absent; content from the project files and tool outputs is processed directly without escaping or validation.

skill-ship