The Agent Skills Directory

[COMMAND_EXECUTION]: Dynamic context injection in SKILL.md executes a local script at load time using the '!' syntax. The inclusion of $ARGUMENTS directly in the shell command presents a risk of command injection if the platform does not sanitize user input.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. 1. Ingestion points: Technical specification documents read by tech_spec_parser.py. 2. Boundary markers: Absent; parsed content is injected directly into the agent's context. 3. Capability inventory: The agent has access to Read, Glob, Grep, and Bash tools. 4. Sanitization: Absent; the parser does not filter for embedded instructions in the spec file.
[COMMAND_EXECUTION]: The parser script uses dynamic module loading via import to load the tree-sitter-markdown library, which is a potential risk factor for executing unintended code.

tech-spec