The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to verify environment variables (GEMINI_API_KEY, COMPOSIO_API_KEY) and the existence of local documentation files. It uses curl to send data to the Google Gemini API and a Python one-liner to parse the JSON response.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to generativelanguage.googleapis.com (Google Gemini API). It also references an installation command using npx from the @opendirectory.dev registry, which is the official namespace for the skill's authoring platform.
[DATA_EXFILTRATION]: Business context, including target personas and account details retrieved from docs/icp.md and account-specific markdown files, is sent to the external Google Gemini API. This is the intended core functionality for personalizing the outreach messages.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests untrusted data from local markdown files and interpolates it into the prompt sent to the LLM.
Ingestion points: Content from docs/icp.md and docs/accounts/*.md is loaded and used to construct the final API request.
Boundary markers: There are no explicit delimiters or 'ignore' instructions implemented to isolate the ingested file content from the system instructions in the prompt template.
Capability inventory: The skill has access to network operations (curl), local execution (python3), and file system writes (cat, mkdir), which could be abused if an attacker can influence the files read by the skill.
Sanitization: No validation or escaping is performed on the content of the markdown files before processing.

outreach-sequence-builder