The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to execute system commands including git, gh (GitHub CLI), and the printing-press binary. These operations are core to the skill's purpose of managing repository state and publishing code.\n- [EXTERNAL_DOWNLOADS]: The setup process in SKILL.md may trigger a download and installation of the printing-press binary via go install github.com/mvanhorn/cli-printing-press/v4/cmd/printing-press@latest. This targets a repository owned by the skill's author.\n- [REMOTE_CODE_EXECUTION]: The validation step in SKILL.md executes go run golang.org/x/vuln/cmd/govulncheck@v1.3.0 ./..., which downloads and runs a security scanning tool from the official Go vulnerability database. This is a standard security practice for Go development.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface (Category 8). It ingests data from local CLI directories, such as .printing-press.json manifests and README.md files, and uses this content to populate pull request descriptions and novel feature tables.\n
Ingestion points: Reads metadata and documentation from local project directories via printing-press library list and direct file reads as described in SKILL.md.\n
Boundary markers: Absent. The skill instructions describe direct interpolation of manifest fields and README excerpts into the PR body template.\n
Capability inventory: The skill has broad capabilities including Bash (git, gh, go, rm, cp), Write, and Edit tool access as defined in SKILL.md.\n
Sanitization: Includes mandatory PII scrubbing and secret scanning (vendor-prefix tokens, gitleaks, trufflehog) to prevent data exposure, but does not specifically sanitize against malicious instructions in the input data that might influence the agent's PR generation logic.

printing-press-publish