The Agent Skills Directory

[SAFE]: The skill is well-structured with defined security boundaries. It includes a proactive redaction system in scripts/discover_assets.py that identifies and masks API keys, bearer tokens, and other sensitive credential patterns before they are passed to the agent context. Additionally, the skill's instructions specifically warn against reading sensitive directories like .ssh or .env.
[COMMAND_EXECUTION]: Subprocess calls in scripts/discover_assets.py are used exclusively to gather git metadata (tags, logs, remotes). The script correctly implements these calls using list-based arguments rather than shell strings, which mitigates command injection risks.
[PROMPT_INJECTION]: The skill was evaluated for indirect prompt injection risks associated with scanning untrusted skill directories.
Ingestion points: scripts/discover_assets.py reads content from files within a user-specified target directory (SKILL.md, documentation, scripts).
Boundary markers: Extracted content is presented to the agent using structured markdown headers (e.g., ### [role] path) and code blocks, which helps maintain context separation.
Capability inventory: The skill has permissions to execute specific local scripts and write to the target directory or /tmp/. It does not have network access tools.
Sanitization: The skill identifies and skips known sensitive file paths (e.g., .ssh, .aws, .env) and uses a redaction engine to strip potential secrets from extracted text.

skill-card-generator