The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides templates for workflows that ingest untrusted data from external sources (HTTP webhooks), which could be leveraged for prompt injection attacks if the data is subsequently passed to an LLM without proper sanitization.
Ingestion points: The steps.trigger.event object in SKILL.md captures incoming webhook data, including bodies and headers.
Boundary markers: The provided code examples do not demonstrate the use of delimiters or specific instructions to ignore embedded commands within the ingested data.
Capability inventory: The skill demonstrates capabilities for network communication via Slack (this.slack.chat.postMessage), external API interaction via the stripe package, and state management via this.db.set.
Sanitization: While the skill mentions signature verification for security, it does not provide examples for sanitizing or escaping content before it reaches downstream components.
[EXTERNAL_DOWNLOADS]: The documentation references the installation of the official Pipedream CLI for local development and deployment.
Evidence: npm install -g @pipedream/cli in SKILL.md targets a well-known service domain for the platform.
[COMMAND_EXECUTION]: The skill instructs the user to execute CLI commands for platform authentication and workflow deployment.
Evidence: pd login and pd deploy workflow.yaml commands are documented in the Installation section of SKILL.md.

pipedream