The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, such as URL response headers and CI log files, to determine the necessary link repair actions.
Ingestion points: playbooks/full-audit.md extracts data from GitHub Action logs (audit/baseline.log) via gh run view. references/tier3-substitution.md probes external URLs using curl and the gh api.
Boundary markers: Absent. The instructions do not define explicit delimiters to separate external data from the agent's logic.
Capability inventory: The skill possesses extensive capabilities including Bash (for git, gh, sed, curl), Write, and Edit permissions.
Sanitization: Absent. The skill lacks a formal sanitization or validation layer for the external content before processing it for file modifications.
[COMMAND_EXECUTION]: The skill uses shell commands to automate repository management and link auditing tasks.
Evidence: playbooks/full-audit.md performs gh repo clone and gh run view. references/tier2-intra-repo.md and references/tier3-substitution.md use sed to modify repository files at scale.
[EXTERNAL_DOWNLOADS]: The skill fetches repository content and verifies external links via network requests.
Evidence: playbooks/full-audit.md clones GitHub repositories. references/tier3-substitution.md uses curl and gh api to verify the existence and status of external resources.

alterlab-link-health