Assess AI Change Impact

SkillAI & models

Assess the direct and transitive impact of a material AI-system or engagement change before promotion. Use when a source, workflow claim, policy, data contract, model route, tool, capability, evaluator, runtime, operation, or user surface changes.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Assess AI Change Impact skill

What this skill tells your AI

The instructions your AI receives, as published by davidahmann/applied-ai-field-guide in .agents/skills/assess-ai-change-impact/SKILL.md and read by ahel’s review.

Use declared dependencies and current source revisions to find what must be reviewed. The map routes inspection; it does not authorize promotion or prove completion.

Read first

  1. Read the evidence graph and change-intelligence blueprint, map freshness and change impact, and change management.
  2. Use the current system-map manifest and change-impact assessment. If no fresh map exists, inspect primary artifacts directly.
  3. Read only the affected solution context; it is not a substitute for target dependencies or release evidence.
  4. Apply CTX-004, DEL-001, DEL-002, OPS-007, and all controls governing the changed boundary.

Workflow

  1. Define the changed subject, prior and proposed revisions or digests, reason, materiality, affected workflow and segment, owner, and required decision.
  2. Verify map scope, producer, source revisions, classification, freshness, coverage, and invalidation triggers. Mark it stale when any binding is missing or changed.
  3. Identify direct dependencies from declared relations and primary references. Traverse transitive dependencies without converting semantic similarity into confirmed impact.
  4. Label each candidate confirmed, inferred, unaffected, or unknown; cite the relation or source and name the owner who must dispose material uncertainty.
  5. Inspect workflow, data, identity, evaluation, operations, economics, user-surface, migration, rollout, rollback, and dependency-lifecycle consequences.
  6. Preserve unaffected revisions and prior decisions. Update only dependency-linked work; route stale or conflicting artifacts for review.
  7. Define per-route tests, representative cases, soak or canary, rollback trigger, source-of-truth verification, and post-change outcome check.
  8. Decide proceed_to_validation, revise_scope, block, or rollback. Incomplete critical scope blocks promotion.

Output contract

Return:

  • a completed change-impact assessment bound to the exact map and change revisions;
  • direct and transitive impacts with evidence, confidence class, owner, and disposition;
  • explicit unaffected and unknown areas;
  • required validation, migration, rollout, rollback, and post-change checks;
  • one promotion decision.

Do not let a graph, generated summary, model inference, merge, or passing subset of tests authorize a release.

Signals

GitHub stars
105
Forks
22
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
assess-ai-change-impact
Source
github.com/davidahmann/applied-ai-field-guide