The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it ingests untrusted data from conversations and the competitive library to generate sales artifacts and emails.
Ingestion points: External data enters the agent context via search_knowledge_base, list_findings, and get_playbook tools as seen in Step 2 of the instructions in SKILL.md.
Boundary markers: The generated templates in Step 3 (e.g., Mode: Full Battlecard) use standard markdown headers, but do not include explicit instructions for the agent to ignore any embedded directives within the retrieved data.
Capability inventory: The skill has the capability to generate content and outreach emails via the generate_content and generate_email MCP tools.
Sanitization: There is no evidence of filtering, escaping, or validation of the retrieved library data before it is processed into the final output.
[DATA_EXFILTRATION]: The skill accesses internal deal data and conversation transcripts using authorized tools (list_events, list_findings). No unauthorized network operations or exfiltration to external domains were identified.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns, external package installations, or dynamic code execution behaviors were found. The skill relies exclusively on defined MCP tools for its functionality.
[COMMAND_EXECUTION]: No arbitrary command execution or use of subprocesses was detected. All operations are performed through structured tool calls.

battlecard