The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to execute npx pxpipe-proxy, which downloads and runs code from the npm registry. This dependency is not from a verified or well-known source.
[COMMAND_EXECUTION]: The core functionality of the skill depends on executing shell commands to process files using the pxpipe-proxy tool.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the content it processes.
Ingestion points: The skill ingests untrusted data from files, directories, git refs, or stdin as specified in SKILL.md.
Boundary markers: The skill lacks instructions for the agent to use delimiters or to ignore potential instructions embedded within the processed text or generated images.
Capability inventory: The agent is instructed to read generated files (page-*.png, factsheet.txt) and perform follow-up actions like Grep, which could be influenced by malicious content in the source data.
Sanitization: No sanitization or filtering is performed on the input data before it is rendered or summarized in the factsheet.

honey-px