The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources and interpolates it into Slack messages, which could allow malicious content to be displayed to users or influence the agent's behavior.
Ingestion points: The skill ingests target account data via CSV file uploads and consumes external event data (news, job postings, and hire information) from Signal payloads as described in SKILL.md.
Boundary markers: Absent. The instructions for constructing Slack messages using markdown templates do not specify the use of delimiters or 'ignore' warnings for the interpolated data.
Capability inventory: The skill uses the clay CLI for resource orchestration and has the capability to write messages to Slack via the send-message-to-channel-botname action in SKILL.md.
Sanitization: Partial. While the skill performs basic normalization on account identifiers (e.g., lowercase, stripping 'www.'), there is no documented sanitization for the job titles, news headlines, or person names that are directly included in the Slack alerts.
[COMMAND_EXECUTION]: The skill relies on executing commands via the clay CLI to interact with the platform's API and manage resources.
Evidence: The instructions in SKILL.md detail the use of clay whoami, clay signals create, clay audiences fields list, and clay workflows publish. These are standard tools provided by the vendor (clay-run) for managing workflows and data segments within their ecosystem.

alert-reps-on-account-signals