The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git commands via subprocess.run to clone repositories and retrieve metadata like commit history and remote URLs. The implementation correctly uses argument lists rather than shell strings, which effectively mitigates shell injection risks.
[EXTERNAL_DOWNLOADS]: Automatically clones external git repositories from GitHub to a local 'upstream' directory based on user-provided URLs. This involves network activity and the ingestion of content from arbitrary external sources.
[PROMPT_INJECTION]: Creates an indirect prompt injection surface by extracting the first sentence of README files from cloned repositories and inserting them into Obsidian project cards. If the agent later reads these cards, malicious instructions embedded in a repository's README could potentially influence the agent's behavior.
Ingestion points: scripts/gen_repo_catalog.py (reads README.md and other documentation from cloned repositories).
Boundary markers: Absent. The extracted summaries are inserted into markdown files without delimiters or instructions to the agent to ignore embedded commands.
Capability inventory: scripts/gen_repo_catalog.py (executes git via subprocess.run with parameters).
Sanitization: Basic text cleaning (stripping HTML tags and Markdown formatting) is performed, but there is no validation or filtering for instructional language or prompt injection patterns.

soia-pkm-clip-repo