The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the installation of the graphifyy package from PyPI. There is a naming discrepancy between the tool's common name ('graphify') and the requested package name ('graphifyy'), which can be a sign of typosquatting or an unofficial distribution.
[COMMAND_EXECUTION]: The skill utilizes graphify install [platform] commands for a wide range of assistants (Claude, Cursor, Gemini, Copilot, etc.). These commands likely modify local application configurations or install hooks/plugins, which constitutes a persistence-like mechanism and environment modification.
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection (Category 8):
Ingestion points: The tool ingests untrusted external data via graphify add <URL> and by scanning local codebase files (code, docs, PDFs, media).
Boundary markers: There are no instructions provided to the agent to use delimiters or ignore instructions embedded within the processed files or URLs.
Capability inventory: The skill has access to Bash, Read, Write, Grep, and Glob tools, and provides Python snippets that utilize the pathlib and graphify modules to generate reports and HTML visualizations.
Sanitization: No sanitization or validation of the ingested external content is mentioned before it is processed into the GRAPH_REPORT.md or graph.json files which the agent subsequently reads.
[ADVERSARIAL_CONTEXT]: There is a mismatch between the provided author context ('akillness') and the repository/package source ('safishamsi/graphify'). The repository and package are hosted under an account that does not align with the vendor identity provided in the system context.

graphify