The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads and processes external content from GitHub issues and pull request bodies, creating a surface for potential instruction injection via attacker-controlled data.
Ingestion points: Data is ingested via gh issue list, gh api graphql, and gh pr list (SKILL.md).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious content within the issue titles or PR bodies.
Capability inventory: The skill possesses the ability to execute git commands (commit, push), modify files (hotfix_tracking.md), and run embedded scripts.
Sanitization: No sanitization or validation of the ingested text is performed before it is used to influence logic or file contents.
[DYNAMIC_EXECUTION]: The skill uses an embedded Python script to handle patch manipulation at runtime.
Evidence: In Step 5f, a Python script is executed via a bash heredoc (`python3
<< 'EOF'`) to split patch files.
Context: The script is used for utility purposes (patch splitting) and does not directly execute untrusted strings as code, though it operates on data fetched from external PRs.
[COMMAND_EXECUTION]: The skill relies extensively on local CLI tools to perform its tasks.
Tools used: Executes gh (GitHub CLI), git, find, sed, and grep to manage the repository and fetch data.
Verification: Includes environment checks for gh version and authentication status before execution.

dot-release-backport-lts