The Agent Skills Directory

[COMMAND_EXECUTION]: The instruction in SKILL.md Step 3 directs the agent to execute a grep command using the "exact error phrase" from user conversation. This creates a risk of command injection if the input contains shell metacharacters such as backticks or dollar-sign parentheses.
[INDIRECT_PROMPT_INJECTION]: The skill stores untrusted data for later use by other automated agents, creating an injection surface. 1. Ingestion points: User conversation extracted in SKILL.md Step 2. 2. Boundary markers: Uses YAML frontmatter but lacks specific safety delimiters for user-provided Markdown content. 3. Capability inventory: Uses Write tool to create files and Bash to execute search commands. 4. Sanitization: Filenames are sanitized in Step 4, but the content of symptoms and solutions is not sanitized before being written to disk or used in shell commands.

fix-reporter