The Agent Skills Directory

[SAFE]: The skill is entirely focused on its stated purpose of historical analysis and document curation. The prompt logic follows best practices for historical thinking (sourcing, corroboration, contextualization) and includes explicit guidance for identifying absences in the historical record (representation audit). No attempts to bypass safety filters or exfiltrate data were found.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text through the existing_sources and central_question inputs.
Ingestion points: SKILL.md prompt references {{existing_sources}} and {{central_question}}.
Boundary markers: Absent. The prompt interpolates inputs directly into the instruction block.
Capability inventory: None. The skill does not use tools for file writes, network requests, or command execution.
Sanitization: Absent.
Conclusion: While an injection surface exists, the risk is negligible because the skill lacks the capabilities to perform any harmful actions. The only potential impact is the generation of misleading educational rationales.
[DATA_EXFILTRATION]: No network operations or sensitive file path access (e.g., .ssh, .env) are present in the skill code or prompt.
[REMOTE_CODE_EXECUTION]: There are no remote code downloads, package installations, or dynamic execution patterns (such as eval or exec) within the skill.

historical-document-set-curator