The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection. The instructions in SKILL.md direct the agent to 'Address each comment' returned by the tool's output. Because these comments are entered by the user in a web UI and processed as instructions, they could be used to steer the agent towards unauthorized actions.\n
Ingestion points: Review comments returned by the linemark server via stdout after the user completes the review in the browser.\n
Boundary markers: Absent; the agent is not instructed to treat the feedback as untrusted data or to ignore embedded instructions.\n
Capability inventory: The agent is given file-write and development capabilities to resolve the issues raised in the comments.\n
Sanitization: No sanitization is performed on the feedback text before it is presented to the agent.\n- [EXTERNAL_DOWNLOADS]: The skill's instructions in SKILL.md execute the review tool directly from GitHub using npx github:gdaybrice/linemark. This results in the download and execution of the author's code at runtime as the intended delivery mechanism.\n- [COMMAND_EXECUTION]: The server.mjs script executes multiple git commands using execSync and execFileSync to retrieve repository data (diffs, logs, etc.). It also executes a system command to open the user's default web browser.

linemark