The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external template definitions retrieved via MCP tools, which can contain untrusted content that might influence agent behavior.
Ingestion points: The agent ingests template definitions, versions, and metadata through the harness_get and harness_search tools as described in SKILL.md.
Boundary markers: There are no instructions or delimiters provided to the agent to treat the content returned by these tools as untrusted or to ignore any instructions embedded within the template YAML/JSON.
Capability inventory: The skill is restricted to read-only tools (harness_list, harness_get, harness_search, harness_describe) and focuses on report generation; it does not have access to tools for file writing, network exfiltration, or shell command execution.
Sanitization: The instructions do not define any validation or sanitization steps for the data retrieved before it is interpolated into the markdown reports defined in references/report-templates.md.

template-usage