The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local shell script (scripts/repo-audit.sh) to perform repository audits. The script follows security best practices such as utilizing set -euo pipefail and validating the existence of the target directory.
[EXTERNAL_DOWNLOADS]: The workflow includes cloning external repositories from GitHub for auditing and analysis. These operations target a well-known service (GitHub) and are consistent with the skill's stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted data from external repositories, creating a surface for indirect prompt injection.
Ingestion points: The agent clones and reads arbitrary files from remote repositories during the audit process in Stages 1 and 2.
Boundary markers: The instructions do not specify explicit boundary markers or directives for the agent to disregard instructions found within the audited files.
Capability inventory: The skill allows the agent to read file contents, execute a bundled audit script, modify repository files (Stage 6), and push updates back to GitHub (Stage 7).
Sanitization: The audit script (scripts/repo-audit.sh) includes robust sanitization logic for its link-checking feature, employing pwd -P and manual path-component inspection to prevent directory traversal and ensure validated links remain within the repository root.

openclaw-github-repo-commander