The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes instructions for the agent to use standard development tools such as git and gh (GitHub CLI) to synchronize project state with remote repositories. These operations are intended for routine project management.
[COMMAND_EXECUTION]: Test and benchmark scripts (tests/run_tests.py, tests/bench.py) utilize subprocess.run to execute internal Python utility scripts for verification. These calls are limited to the skill's own codebase and operate within temporary directories or the project's designated memory root.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes external data from remote git logs and pull request metadata. It mitigates this risk through explicit instructions that command the agent to manually review, paraphrase, and never treat remote wording as authoritative conclusions.
Ingestion points: Markdown files located in the ./tmp/remote/ directory, which are populated from external repository data.
Boundary markers: Instructional safeguards in SKILL.md and references/remote-sync.md requiring the agent to distinguish between remote content and local project state.
Capability inventory: The agent has the capability to read/write files and execute shell commands (git, gh) based on the data processed.
Sanitization: The workflow requires the agent to rewrite and validate all findings before they are promoted to stable project documentation.

project-memory