The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources and interpolates it into agent queries. Specifically, it uses company names retrieved from Clari meeting logs to query deal status (e.g., 'Does [company] have an open deal in Sales Pipeline?'). 1. Ingestion points: Google Drive spreadsheet content and Clari meeting summaries via clari_list_calls. 2. Boundary markers: No delimiters or safety instructions are provided when interpolating external data into queries. 3. Capability inventory: The skill has permissions to write to Google Drive, post to Slack, and perform network requests to the Epiphan AI API. 4. Sanitization: No sanitization or validation of the ingested strings is performed before they are used in prompts.
[DATA_EXFILTRATION]: Sensitive business metrics, including weekly and YTD revenue, deal counts, and pipeline values, are extracted from internal systems and written to a local file at ~/.claude/skill-analytics/last-outcome-epiphan-weekly-sales-tracker-skill.json. Storing these metrics in a predictable local path outside the repository scope could lead to data exposure.
[EXTERNAL_DOWNLOADS]: The skill configuration relies on npx to dynamically fetch and run the mcp-remote package from the npm registry at runtime. While npm is a standard registry, this creates a dependency on external code that is executed during the initialization of the MCP server.
[COMMAND_EXECUTION]: The skill executes shell commands via npx to launch the mcp-remote client, which is used to communicate with the remote Epiphan AI MCP server.

epiphan-weekly-sales-tracker-skill