The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to run local system commands such as git, find, and grep to audit project activity. These operations are restricted to the local environment and are used to gather metadata for the improvement report.
[PROMPT_INJECTION]: The skill implements an audit loop that reads local project documentation (e.g., CLAUDE.md, MEMORY.md) to generate summaries and skill drafts. This represents an indirect prompt injection surface where content in those files could influence agent output, though the risk is localized to the user's own workspace.
Ingestion points: Phase 3 reads several context-providing markdown files (CLAUDE.md, MEMORY.md, etc.) and recently modified files in the working directory.
Boundary markers: The skill does not explicitly use boundary markers or "ignore instructions" warnings when processing the ingested file content.
Capability inventory: The skill has access to the Bash, Write, and Edit tools, which it uses to manage local files and generate the audit report and skill drafts.
Sanitization: There is no evidence of explicit sanitization or escaping of the content read from the local files before it is used in prompt generation.

weekly-setup-improvements