The Agent Skills Directory

[SAFE]: The skill is strictly limited to generating Markdown documentation. It explicitly instructs the agent not to implement code, stage files, or perform git operations.
[DATA_EXPOSURE]: The skill reads conversation history and repository structure to generate context for the handoff plan. This is standard behavior for a summarization tool and does not involve accessing sensitive credentials or system files.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data to generate documentation. 1. Ingestion points: Current conversation history and repository file content (SKILL.md, Step 2). 2. Boundary markers: None explicitly defined to separate instructions from user-provided context. 3. Capability inventory: Filesystem write access to create or update Markdown files. 4. Sanitization: No sanitization or filtering of the ingested conversation data is performed. The risk is considered safe as the output is restricted to non-executable Markdown and the instructions prohibit implementation actions.

progress-handoff