The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from various external sources, which creates a surface for indirect prompt injection attacks.
Ingestion points: SKILL.md (Workflow steps 3 and 4) and references/demand-evidence-sources.md instruct the agent to pull data from docs-site search logs, support tickets, and issue trackers.
Boundary markers: The instructions do not define explicit delimiters (e.g., XML tags or special characters) to wrap the untrusted content or provide "ignore embedded instructions" warnings for the agent.
Capability inventory: The skill is composed of instructions and references; it does not contain code or scripts with direct file-system access, network operations, or shell execution capabilities.
Sanitization: Step 4 in the SKILL.md workflow explicitly requires normalizing error strings by replacing absolute paths, UUIDs, IP addresses, and numbers with placeholders.
[SAFE]: The skill incorporates strong privacy practices by requiring the agent to scrub and normalize sensitive identifiers from external logs and tickets before they are clustered or included in reports.
[SAFE]: All external URLs and references (e.g., Ahrefs, SparkToro, Google Search Central, Stack Overflow, and JetBrains) target official documentation and research from well-known and trusted organizations.
[SAFE]: The skill correctly scopes its behavior and references other tools within the same vendor namespace ('samber'), representing normal cross-skill integration.

developer-keyword-research