The Agent Skills Directory

[DATA_EXFILTRATION]: The skill implements a --deliver webhook:<url> flag across its command suite, providing a built-in mechanism for the agent to exfiltrate research results, account metadata, and financial data to arbitrary external URLs.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface through its local 'learning' and 'playbook' system. The agent is instructed to follow next_action and playbook steps retrieved from a local SQLite database that records previous research activity; this data could be poisoned by malicious content encountered during web research, leading to the execution of unintended commands.
Ingestion points: Output from parallel-pp-cli recall as described in Step 2 of the decision tree.
Boundary markers: Absent; the agent is explicitly told to follow candidates and playbooks 'verbatim'.
Capability inventory: The CLI can perform financial transactions (account-add-balance), manage API keys, and access local research history.
Sanitization: No sanitization or validation of the retrieved instructions is documented.
[COMMAND_EXECUTION]: The skill facilitates high-risk administrative and financial operations, including charging payment methods (service account-add-balance) and generating/deleting API credentials (service account-create-key).
[EXTERNAL_DOWNLOADS]: The skill installs its core binary from official vendor repositories on GitHub (github.com/mvanhorn/printing-press-library) and NPM (@mvanhorn/printing-press-library).

pp-parallel