The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard system utilities and CLI tools including git, gh, curl, rg, shasum, and mktemp. These are used appropriately to determine project context, create structured local files, and interact with GitHub for Gist delivery.
[DATA_EXFILTRATION]: The skill facilitates the transfer of local artifact content to GitHub Gists. This is a primary intended feature and is protected by multiple security guardrails: it requires explicit human approval for every upload, enforces private visibility (public=false), and includes a dedicated verification script (verify-gist-delivery) that performs a regex scan for sensitive patterns such as RSA/OPENSSH private keys and machine-local absolute paths (/Users/, /home/, /nix/store/).
[INDIRECT_PROMPT_INJECTION]: The skill processes content from local markdown files and reads metadata/raw content back from GitHub for verification. While it lacks formal boundary markers for untrusted data, the processing logic is limited to integrity checks (byte comparison, SHA-256 hashing) and security scanning, which does not involve interpreting the file content as instructions.
[SAFE]: The provided Bash scripts demonstrate defensive coding practices, including strict error handling (set -o errexit, nounset, pipefail) and validation of user-supplied directory/label components to prevent path traversal. No indicators of obfuscation, persistence, or privilege escalation were found.

logbook