The Agent Skills Directory

[SAFE]: The skill instructions and reference materials consistently prioritize security, specifically advising against hardcoding credentials and requiring human intervention for actions that impact accounts or persistent state.
[INDIRECT_PROMPT_INJECTION]: The skill operates on an inherent attack surface by ingesting external data (e.g., scraped web content, emails, and reviews) for marketing analysis.
Ingestion points: External data enters the agent context via sources described in SKILL.md and references/skill-contract.md.
Boundary markers: The skill includes clear directives to ignore any instructions embedded in analyzed data and to treat such content strictly as evidence.
Capability inventory: The instructions coordinate analysis that potentially informs actions in tools like Slack, HubSpot, and advertising platforms.
Sanitization: The risk is mitigated through instructional guardrails and a structured proposal-approval lifecycle rather than programmatic code-based filtering.
[NO_CODE]: The skill contains only documentation, instruction sets, and reference schemas in Markdown format; no executable scripts, binaries, or package dependencies are included.

marketing-loops