fpf-skillstorage-persist-evidence

venikman's avatarfrom venikman

Writes an immutable artifact to the FPF EvidenceGraph (G.6).

⭐0stars🔀0forks📁View on GitHub🕐Updated Jan 10, 2026
[Agent Observability]

When & Why to Use This Skill

This Claude skill facilitates the creation of immutable data artifacts within the FPF EvidenceGraph framework. It ensures that critical telemetry and process evidence are stored securely with unique, collision-resistant filenames (SHA256/UUID), providing a reliable and non-repudiable audit trail for agentic workflows and automated experiments.

Use Cases

  • Audit Logging: Recording step-by-step actions and decisions of an AI agent to ensure full accountability and traceability in production environments.
  • Experiment Documentation: Saving immutable snapshots of data during research phases to prevent accidental overwrites and maintain data integrity.
  • Telemetry Collection: Systematically storing execution logs and performance metrics in a structured JSON format for downstream analysis and monitoring.
  • Compliance and Verification: Generating verifiable evidence of automated tasks to meet strict regulatory requirements or quality assurance standards.
namefpf-skill:storage-persist-evidence
descriptionWrites an immutable artifact to the FPF EvidenceGraph (G.6).
version0.1.0
allowed_tools[]

storage/persist-evidence Kernel

Context

This skill implements FPF Pattern G.6 (EvidenceGraph). In the experiment (Conditions C1/C2), files act as the carrier. Evidence must be immutable and addressable.

Instructions

Write the provided content to the telemetry/ or artifacts/ directory.

1. Naming Convention (Immutability)

You must generate a filename that guarantees uniqueness/immutability.

  • Preferred: timestamp_sha256.json (e.g., 20240110-1200_a1b2c3...json)
  • Alternate: uuid.json

2. Formats

  • Content MUST be valid JSON.
  • If content is not JSON, wrap it: {"content_type": "text/plain", "body": "..."}.

3. Output

Return the absolute path of the persisted artifact.