The Agent Skills Directory

[SAFE]: The skill is primarily instructional, offering templates and reference implementations for risk management workflows. The included JavaScript and Python snippets in the references/ directory are self-contained logic for calculating risk scores and organizing risk registers, with no external network or system-level interactions.\n- [INDIRECT_PROMPT_INJECTION]: The skill contains logic to process external inputs such as project context and assumptions. While these ingestion points could theoretically be used for indirect prompt injection, the skill does not possess any dangerous capabilities (e.g., file system writes, network requests, or command execution) that would make such an injection actionable. The provided code is limited to data transformation and display logic.\n
Ingestion points: The analyze_assumptions_as_risks and brainstorm_risks functions in references/risk-identification-techniques.md accept external data as arguments.\n
Boundary markers: None present in the sample code to delimit or instruct the model to ignore embedded commands.\n
Capability inventory: No network access, subprocess execution, or file write capabilities were found across any of the referenced files.\n
Sanitization: No explicit sanitization or validation of the input strings is present in the illustrative code snippets.

risk-assessment