The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill attempts to install a Python package named graphifyy via pip. The name differs from the skill name graphify, which is a common indicator of a non-standard or potentially suspicious dependency.
[PERSISTENCE]: The skill includes multiple mechanisms to maintain persistence and influence future agent behavior:
The graphify hook install command installs a git post-commit hook that executes code automatically after every commit.
The native Agent integration instructions append rules to CLAUDE.md, GEMINI.md, and AGENTS.md to permanently alter how agents interact with the project.
The --watch flag starts a background process to monitor file changes and trigger re-extractions.
[INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary data from a variety of sources, creating a significant attack surface.
Ingestion points: Processes any files in a folder (code, docs, papers, images, videos) and fetches remote content from URLs via the /graphify add command.
Boundary markers: The instructions for subagents lack explicit delimiters or "ignore embedded instructions" warnings, increasing the risk that the subagent will follow instructions hidden in the processed data.
Capability inventory: Subagents are explicitly required to be general-purpose, granting them access to Write and Bash tools. The main skill also performs file system writes and network operations (ingest, push_to_neo4j).
Sanitization: There is no evidence of sanitization, filtering, or validation of the content ingested from local files or external URLs before it is processed by the AI.
[COMMAND_EXECUTION]: The skill performs extensive shell command execution to manage the environment, including detecting Python interpreters, installing packages with --break-system-packages, and writing configuration files to the file system.

graphify