Appendix: clarity.baseline@1.0

Status: Normative Appendix
Last Updated: 2026-02-28
Related Specs: ACGP-1, ACGP-3

This appendix defines the normative baseline blueprint clarity.baseline@1.0 used as the mandatory root blueprint for v1.0 inheritance chains.

1. Normative Baseline Requirements

clarity.baseline@1.0 MUST include:

  1. No tripwires
  2. Default GT-0-aligned risk thresholds using closed-upper-bound semantics:
  3. ok if risk ≤ 0.40
  4. nudge if 0.40 < risk ≤ 0.55
  5. escalate if 0.55 < risk ≤ 0.70
  6. block if risk > 0.70 (implicit; 3-threshold model)
  7. All five standard CTQ metrics expressed as checks[] entries with equal weights (0.20 each)
  8. No evidence requirements
  9. Trust debt enabled with default accumulation/decay rates, including flag: 0.1

2. YAML Representation (Normative)

artifact_type: acgp.blueprint
schema_version: "2.0.0"
id: clarity.baseline@1.0
version: "1.0.0"
title: "Clarity Baseline"
description: "Universal cognitive governance baseline for ACGP v1.0"
checks:
  - id: reasoning_quality
    kind: metric
    metric:
      name: reasoning_quality
      weight: 0.20
      evaluator:
        kind: rule-based
  - id: knowledge_grounding
    kind: metric
    metric:
      name: knowledge_grounding
      weight: 0.20
      evaluator:
        kind: rule-based
  - id: ethical_alignment
    kind: metric
    metric:
      name: ethical_alignment
      weight: 0.20
      evaluator:
        kind: rule-based
  - id: tool_safety
    kind: metric
    metric:
      name: tool_safety
      weight: 0.20
      evaluator:
        kind: rule-based
  - id: context_awareness
    kind: metric
    metric:
      name: context_awareness
      weight: 0.20
      evaluator:
        kind: rule-based
tripwires: []

intervention_policy:
  thresholds:
    ok: 0.40
    nudge: 0.55
    escalate: 0.70

trust_policy:
  enabled: true
  accumulation:
    ok: 0.0
    flag: 0.1
    nudge: 0.5
    escalate: 1.0
    block: 2.0
    halt: 5.0
  decay:
    decay_fraction: 0.05
    period_hours: 1
    min_debt: 0.0
  thresholds:
    elevated_monitoring: 3.0
    restricted_mode: 6.0
    re_tiering_review: 10.0

Note: v1.0 blueprints express CTQ metrics through the checks[] array. There is no separate ctq top-level field in v1.0. Implementations MUST NOT expect or require a standalone ctq block.

3. JSON Representation (Normative)

{
  "artifact_type": "acgp.blueprint",
  "schema_version": "2.0.0",
  "id": "clarity.baseline@1.0",
  "version": "1.0.0",
  "title": "Clarity Baseline",
  "description": "Universal cognitive governance baseline for ACGP v1.0",
  "checks": [
    { "id": "reasoning_quality", "kind": "metric", "metric": { "name": "reasoning_quality", "weight": 0.20, "evaluator": { "kind": "rule-based" } } },
    { "id": "knowledge_grounding", "kind": "metric", "metric": { "name": "knowledge_grounding", "weight": 0.20, "evaluator": { "kind": "rule-based" } } },
    { "id": "ethical_alignment", "kind": "metric", "metric": { "name": "ethical_alignment", "weight": 0.20, "evaluator": { "kind": "rule-based" } } },
    { "id": "tool_safety", "kind": "metric", "metric": { "name": "tool_safety", "weight": 0.20, "evaluator": { "kind": "rule-based" } } },
    { "id": "context_awareness", "kind": "metric", "metric": { "name": "context_awareness", "weight": 0.20, "evaluator": { "kind": "rule-based" } } }
  ],
  "tripwires": [],
  "intervention_policy": {
    "thresholds": {
      "ok": 0.40,
      "nudge": 0.55,
      "escalate": 0.70
    }
  },
  "trust_policy": {
    "enabled": true,
    "accumulation": {
      "ok": 0.0,
      "flag": 0.1,
      "nudge": 0.5,
      "escalate": 1.0,
      "block": 2.0,
      "halt": 5.0
    },
    "decay": {
      "decay_fraction": 0.05,
      "period_hours": 1,
      "min_debt": 0.0
    },
    "thresholds": {
      "elevated_monitoring": 3.0,
      "restricted_mode": 6.0,
      "re_tiering_review": 10.0
    }
  }
}

4. End-to-End Alpha.2 Example (Informative)

This example shows how the mandatory clarity.baseline@1.0 root participates in an alpha.2 Standard-profile evaluation flow. It is informative, but it is designed to mirror the current normative claim surface and the shipped conformance vectors.

4.1 Effective Blueprint Shape

An implementation starts from clarity.baseline@1.0 and layers a workload blueprint on top. A minimal child blueprint might look like this:

artifact_type: acgp.blueprint
schema_version: "2.0.0"
id: payments.standard@1.0
version: "1.0.0"
title: "Payments Standard"
description: "Workload blueprint layered on the mandatory clarity baseline."
base:
  ref: clarity.baseline@1.0

applicability:
  governance_tiers: [GT-3]

checks: []
tripwires:
  - id: high_value_transfer
    when:
      hook: tool_call
    condition: "args.amount > 10000"
    on_fail:
      decision: block
      reason: "Transfer exceeds automated approval band"

intervention_policy:
  thresholds:
    ok: 0.40
    nudge: 0.55
    escalate: 0.70

The effective runtime blueprint therefore includes:

  1. The Clarity Baseline CTQ checks and thresholds.
  2. Trust debt enabled with the default alpha.2 accumulation and fractional decay semantics.
  3. One workload-specific tripwire evaluated against the canonical args root.

4.2 Trace Chain

The agent emits a parent trace for the session and a child trace for the actual governed tool call:

{
  "trace_id": "trace-root-001",
  "hook": "session_init",
  "reasoning": "Open a governed transfer workflow for the authenticated customer.",
  "action": {
    "name": "start_session",
    "parameters": {
      "channel": "web",
      "workflow": "funds_transfer"
    }
  },
  "governance_tier": "GT-3"
}
{
  "trace_id": "trace-child-002",
  "parent_trace_id": "trace-root-001",
  "hook": "tool_call",
  "reasoning": "Transfer request is within the local approval band and can be evaluated automatically.",
  "action": {
    "name": "transfer_funds",
    "parameters": {
      "amount": 2500,
      "currency": "USD",
      "destination": "acct-77"
    }
  },
  "governance_tier": "GT-3"
}

This demonstrates the minimal lineage rule from ACGP-1: a child trace MAY link to its immediate parent using parent_trace_id, without requiring a full call graph representation in the base protocol.

4.3 Evaluation Outcome

At runtime, the implementation evaluates the child trace in the normative order:

  1. Resolve inheritance and validate the merged blueprint.
  2. Evaluate tripwires first. The expression args.amount > 10000 reads from the canonical args alias for action.parameters.
  3. If no tripwire fires, evaluate rule-based checks and CTQ metrics.
  4. Determine the primary intervention, then apply any orthogonal flag state.
  5. Update trust debt using the final primary intervention plus any flag contribution.
  6. Emit machine-readable evaluation and observability artifacts.

For the child trace above, the transfer amount is below the tripwire threshold, so the expected outcome is a normal Standard-path evaluation, for example:

{
  "trace_id": "trace-child-002",
  "parent_trace_id": "trace-root-001",
  "blueprint_id": "payments.standard@1.0",
  "ctq_dimensions": {
    "reasoning_quality": {
      "score": 0.92,
      "weight": 0.20,
      "status": "evaluated",
      "contributors": ["reasoning_quality"]
    },
    "knowledge_grounding": {
      "score": 0.89,
      "weight": 0.20,
      "status": "evaluated",
      "contributors": ["knowledge_grounding"]
    },
    "ethical_alignment": {
      "score": 0.94,
      "weight": 0.20,
      "status": "evaluated",
      "contributors": ["ethical_alignment"]
    },
    "tool_safety": {
      "score": 0.90,
      "weight": 0.20,
      "status": "evaluated",
      "contributors": ["tool_safety"]
    },
    "context_awareness": {
      "score": 0.91,
      "weight": 0.20,
      "status": "evaluated",
      "contributors": ["context_awareness"]
    }
  },
  "ctq_score": 0.9120,
  "risk_score": 0.0880,
  "tripwires_triggered": [],
  "intervention": "ok",
  "flagged": false,
  "runtime_posture": "normal",
  "review_required": false,
  "trust_debt": {
    "provider_id": "acgp.core.default@1",
    "pre": 0.0,
    "delta": 0.0,
    "post": 0.0,
    "thresholds_crossed": []
  },
  "governance_tier": "GT-3"
}

If the same trace were emitted with amount: 15000, the tripwire would fail closed with primary decision block, the implementation would not execute the transfer, and trust debt would increase according to the baseline accumulation table.

4.4 Implementation Notes

  • This appendix example is informative; conformance is still determined by ACGP-6 and the vector manifest.
  • args is the only shorthand alias for action parameters in tripwire conditions. Other undeclared roots MUST be rejected.
  • parent_trace_id is optional, but when present it should refer to the immediate parent trace only.
  • Safety-Critical is intentionally not used in this example.