The Agent Skills Directory

[SAFE]: The skill utilizes local file system operations (Read, Write, Glob, Grep) to manage session state. These operations are scoped to the current project directory and are consistent with the skill's stated purpose.
[SAFE]: Command execution is limited to basic Bash commands for directory management (mkdir -p) and file organization (mv). No arbitrary command execution or privilege escalation patterns were identified.
[PROMPT_INJECTION]: The skill exhibits an architectural surface for indirect prompt injection by processing external data to generate summaries. However, this is inherent to its summarization function.
Ingestion points: The skill reads CLAUDE.md, project files in ref/ and wip/ directories, and retrieves the last 15-20 messages from the conversation history (SKILL.md).
Boundary markers: No explicit delimiters or 'ignore' instructions are used to separate ingested content from the synthesis logic.
Capability inventory: The skill can perform file creation (Write), file system organization (Bash), and interactive user queries (AskUserQuestion).
Sanitization: Content is synthesized into the final context file without specific filtering or escaping mechanisms described in the instructions.

save-work