The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/check-version.py script includes an --update flag that downloads data from https://raw.githubusercontent.com/dmccreary/claude-skills/main/skills/microsim-matcher/references/matching-criteria.md and overwrites the local references/matching-criteria.md file. Because this repository is not managed by a trusted organization or the skill author, this mechanism allows a remote party to modify the skill's reference data.
[COMMAND_EXECUTION]: The skill workflow in SKILL.md (Step 0) instructs the agent to execute multiple shell commands. These include stat for file metadata, curl to fetch commit information from the GitHub API, and the python interpreter to run the check-version.py script. The execution of these commands provides a baseline capability for system interaction.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection.
Ingestion points: The agent ingests untrusted data from the user's diagram specification (Step 1) and the potentially remote-controlled matching-criteria.md file (Step 3).
Boundary markers: The instructions do not define boundary markers or clear delimiters to separate these data sources from the agent's operational logic.
Capability inventory: The skill has the capability to execute shell commands (stat, curl, python) as seen in Step 0.
Sanitization: There is no evidence of sanitization or validation of the input data before it is used to calculate scores and generate recommendations in Steps 4 and 6.

microsim-matcher