The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection as it ingests and processes untrusted data from web pages.
Ingestion points: Web content is extracted and added to the agent's context through the pinchtab snap and pinchtab text tools as described in SKILL.md and references/api.md.
Boundary markers: While the references/mcp.md file explicitly warns that page-derived output may contain hostile instructions, the skill does not mandate the use of strict delimiters or boundary markers for the ingested content.
Capability inventory: The agent has access to highly capable tools including click, fill, eval (JavaScript execution), download, and upload as detailed in SKILL.md and references/api.md.
Sanitization: There is no evidence of sanitization or filtering of the extracted web content before it is interpolated into prompts.
[COMMAND_EXECUTION]: The skill facilitates the execution of scripts and system-level operations.
The pinchtab eval command allows for the execution of arbitrary JavaScript within the browser context, which is a form of dynamic code execution.
The pinchtab daemon install command is used to install the PinchTab server as a persistent system service, which involves modifying system configuration.
[EXTERNAL_DOWNLOADS]: The skill configuration directs the environment to download the PinchTab binary from an external source.
The SKILL.md frontmatter includes installation instructions for fetching the tool from the pinchtab GitHub repository via brew or go package managers.

pinchtab