The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes external data from GitHub issues to drive its development workflow, creating a surface for indirect prompt injection.
Ingestion points: Fetches issue title, body, and labels using gh issue view <issue_number> in SKILL.md.
Boundary markers: Absent; there are no delimiters or specific instructions to ignore embedded commands within the fetched issue content.
Capability inventory: The skill has broad capabilities including shell command execution (gh, git, npm), file modification (implementation phase), and access to local credentials (Keychains and .env files).
Sanitization: Absent; the agent is instructed to directly "Classify the issue" and "Implement the fix or feature" based on the untrusted input from GitHub.
[DATA_EXFILTRATION]: The skill is configured to access sensitive local files and credential stores to facilitate automated testing.
Evidence: The workflow in SKILL.md explicitly instructs the agent to source "$HOME/.config/taledraw-test.env" if it exists and to retrieve values from the MacOS Keychain using identifiers like codex-talelens-test-email and codex-talelens-test-password.
[COMMAND_EXECUTION]: The skill utilizes the system shell to perform git operations, GitHub CLI interactions, and project-specific build or test scripts.
Evidence: Execution of gh auth status, gh issue view, git push, and various quality check commands such as lint, tsc, and playwright.

gh-issue-autodev