The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/ml-pattern-detection.sh employs the eval command to dynamically load variables from temporary feature files (structural_features.txt, content_features.txt, quality_features.txt). These files are populated by piping outputs from grep, wc, and awk. While the analysis logic currently generates numeric counts, the use of eval on data derived from external files is an insecure practice that could lead to command injection if the analysis output is ever compromised.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content from other skills' SKILL.md files for scoring and similarity analysis.
Ingestion points: Content is ingested into the agent context via scripts/evaluate.sh, scripts/detect-duplication-enhanced.sh, scripts/semantic-analysis.sh, and scripts/ml-pattern-detection.sh.
Boundary markers: Absent; the tool reads and processes raw markdown files directly.
Capability inventory: The skill can read local files, write reports to the .context/ directory, and execute shell utilities like grep, sed, awk, and jq.
Sanitization: Absent; there is no filtering or escaping of the ingested markdown content before it is processed by shell scripts.

skill-quality-auditor