The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install external binaries using npx -y @mvanhorn/printing-press and go install github.com/mvanhorn/printing-press-library/.... These commands download and execute code from third-party repositories (NPM and GitHub).
[DATA_EXFILTRATION]: The CLI tool supports a --deliver webhook:<url> flag, which allows the agent to POST command output (which may include recipe data or local file contents) to an arbitrary external URL. This represents a potential exfiltration vector if the agent is manipulated into sending sensitive information.
[COMMAND_EXECUTION]: The skill is designed to execute the allrecipes-pp-cli binary with high autonomy (using the --agent flag which implies --yes and --no-input), allowing for extensive shell-based operations on the local system.
[INDIRECT_PROMPT_INJECTION]: The tool fetches and processes external recipe data from Allrecipes.com and articles. This creates a surface for indirect prompt injection where malicious instructions embedded in recipe content could influence the agent's behavior during data processing or cookbook generation.
Ingestion points: Recipe pages and articles fetched via recipes get, recipes search, and article <url>.
Boundary markers: Not explicitly defined in the prompt instructions for data handling.
Capability inventory: The skill can execute shell commands via allrecipes-pp-cli, write files via the --deliver file:<path> sink, and perform network requests via the webhook sink.
Sanitization: No explicit sanitization or validation of the external recipe data is mentioned before it is rendered or compiled into cookbooks.

pp-allrecipes