The Agent Skills Directory

[PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by consuming 'Before/After' HTML snippets from the external 'slop-off' MCP server and applying them to the source code.
Ingestion points: External reports retrieved through the wait_for_report, get_latest_report, and list_reports tools (documented in SKILL.md).
Boundary markers: Absent. No delimiters or instructions are provided to the agent to treat the 'After' content as untrusted or to ignore embedded instructions.
Capability inventory: The agent can perform file system writes (modifying project source code) and execute shell commands (typecheck/lint) based on the instructions in SKILL.md.
Sanitization: Absent. The instructions mandate replacing existing code with the 'After-inhoud' content without validation or security scanning of the new code.
[COMMAND_EXECUTION]: The skill directs the agent to execute 'quick checks' such as typecheck or lint after processing a report. This involves spawning subprocesses within the local development environment.

apply-edits