{
  "pack_id": "seeded-defect-and-proof-carrying-review",
  "title": "Seeded defect and proof-carrying review",
  "version": "0.1.0",
  "summary": "A collection of nouns and verbs for mutation-guided reviewer calibration across six seeded defect classes, proof-carrying refusal and approval verification and exact Wilson score plus paired bootstrap confidence intervals.",
  "dependencies": [
    "configuration-grids-and-bounds",
    "checks-of-checks",
    "evaluation-design",
    "wiring-and-conformance-checks"
  ],
  "boundaries": [
    "SeededDefectMutantSuite injects six deterministic defect classes (digest_mismatch, boundary_off_by_one, silent_exception_swallow, capability_tier_violation, self_approval_principal_reuse and ungrounded_citation_span) into candidate bundles.",
    "ProofCarryingReviewVerifier enforces no refusal without a reproducible failing check and witness, and no approval unless reviewer mutant kill rate meets the grid threshold.",
    "StatisticalIntervalAuditor computes exact Wilson score intervals and seeded paired bootstrap confidence intervals from default_grid_spec.json bounds without inline magic numbers.",
    "Importing the pack writes nothing, enqueues nothing and installs no code."
  ],
  "nouns": [
    {
      "slug": "pack",
      "category": "noun.skill_pack",
      "role": "pack",
      "title": "A collection of nouns and verbs for mutation-guided reviewer calibration across six seeded defect classes, proof-carrying refusal and approval verification and exact Wilson score plus paired bootstrap confidence intervals",
      "value": {
        "lets_workers": "Calibrate reviewers against clean-vs-mutant bundle twins across six universal defect classes, reject unfounded refusals that lack a reproducible failing check witness, reject rubber-stamp approvals below mutant kill thresholds and bound pass/kill rates and system deltas with Wilson score and paired bootstrap confidence intervals.",
        "flow": [
          {
            "verb": {
              "$verb": "seeded-defect-and-proof-carrying-review.generate-seeded-defect-mutant-twins"
            },
            "when": "Synthesize deterministic seeded-defect twins across six universal failure classes from clean candidate bundles."
          },
          {
            "verb": {
              "$verb": "seeded-defect-and-proof-carrying-review.verify-proof-carrying-review-verdicts"
            },
            "when": "Calibrate reviewer sensitivity and specificity on clean-vs-mutant twins and reject unfounded refusals or rubber-stamp approvals."
          },
          {
            "verb": {
              "$verb": "seeded-defect-and-proof-carrying-review.audit-wilson-and-paired-bootstrap-intervals"
            },
            "when": "Independently compute Wilson score and paired bootstrap confidence intervals and verify ZeroHardcodedValuesAstGate conformance."
          }
        ],
        "guardrails": [
          {
            "$pin": "guardrail.no-refusal-without-reproducible-witness"
          },
          {
            "$pin": "guardrail.no-approval-without-mutant-kill-calibration"
          }
        ],
        "tool": {
          "$pin": "tool.proof-carrying-review"
        },
        "worked_example": "Two clean candidate bundles expanded into 12 seeded-defect mutants across all 6 defect classes, calibrating a proof-carrying reviewer (1.0 kill rate, Wilson 95% CI [0.7575, 1.0], 0.0 clean FPR), rejecting unfounded refusals and rubber-stamp approvals and computing a paired bootstrap CI [0.215, 0.2263] in test_proof_carrying_review.py."
      },
      "relations": [
        [
          "requires_noun",
          "guardrail.no-refusal-without-reproducible-witness"
        ],
        [
          "requires_noun",
          "guardrail.no-approval-without-mutant-kill-calibration"
        ],
        [
          "requires_noun",
          "spec.seeded-defect-and-review-proof-grid"
        ],
        [
          "requires_noun",
          "rules.mutant-calibration-and-interval-audit"
        ],
        [
          "requires_noun",
          "tool.proof-carrying-review"
        ],
        [
          "requires_noun",
          "template.proof-carrying-review-receipt"
        ]
      ]
    },
    {
      "slug": "guardrail.no-refusal-without-reproducible-witness",
      "category": "noun.skill_pack",
      "role": "guardrail",
      "title": "Guardrail: review refusals must cite at least one reproducible failing check ID and matching witness",
      "category_reason": "Eliminates hallucinated, nitpicky or ungrounded reviewer blocks on clean candidate artifacts.",
      "value": {
        "rule": "ProofCarryingReviewVerifier rejects any review refusal that fails to name an actual failing deterministic check_id and witness reproduced against the candidate bundle.",
        "how_it_is_checked": "ProofCarryingReviewVerifier.verify_review_decision re-executes SeededDefectMutantSuite.run_checks and rejects empty or non-reproducing refusal claims."
      }
    },
    {
      "slug": "guardrail.no-approval-without-mutant-kill-calibration",
      "category": "noun.skill_pack",
      "role": "guardrail",
      "title": "Guardrail: review approvals require calibrated mutant kill rates and zero inline magic literals",
      "category_reason": "Prevents rubber-stamp approvals and ensures statistical confidence parameters resolve strictly from default_grid_spec.json.",
      "value": {
        "rule": "ProofCarryingReviewVerifier rejects approvals when underlying deterministic checks fail or when reviewer mutant_kill_rate falls below min_mutant_kill_rate in default_grid_spec.json, and ZeroHardcodedValuesAstGate rejects inline magic numbers.",
        "how_it_is_checked": "test_proof_carrying_review.py verifies rubber-stamp rejection, 6-class mutant detection, exact Wilson/bootstrap intervals and zero AST magic-literal violations."
      }
    },
    {
      "slug": "spec.seeded-defect-and-review-proof-grid",
      "category": "noun.benchmark_suite",
      "role": "spec",
      "title": "Specification: six universal seeded defect classes, proof-carrying verdict contract and interval estimators",
      "category_reason": "Defines deterministic mutant operators, proof-carrying review invariants and statistical interval equations.",
      "value": {
        "structures": [
          "SeededDefectMutantSuite",
          "ProofCarryingReviewVerifier",
          "StatisticalIntervalAuditor"
        ],
        "defect_classes": [
          "digest_mismatch",
          "boundary_off_by_one",
          "silent_exception_swallow",
          "capability_tier_violation",
          "self_approval_principal_reuse",
          "ungrounded_citation_span"
        ],
        "verdict_invariants": [
          "no_refusal_without_failing_check",
          "no_approval_without_mutant_kill_threshold"
        ],
        "interval_estimators": [
          "wilson_score_interval",
          "paired_bootstrap_interval"
        ]
      }
    },
    {
      "slug": "rules.mutant-calibration-and-interval-audit",
      "category": "noun.reflex_procedure",
      "role": "rules",
      "title": "Deterministic twin-mutant reviewer calibration, witness verification and confidence interval rules",
      "category_reason": "Zero-model procedure scoring reviewer sensitivity/specificity and statistical confidence bounds.",
      "value": {
        "wilson_equation": "(p_hat + z^2/(2n) +- z * sqrt((p_hat*(1 - p_hat) + z^2/(4n))/n)) / (1 + z^2/n)",
        "paired_delta_definition": "candidate_score[i] - baseline_score[i]",
        "model_calls": 0
      }
    },
    {
      "slug": "tool.proof-carrying-review",
      "category": "noun.code_node",
      "role": "tool_card",
      "title": "Tool card: standard-library SeededDefectMutantSuite, ProofCarryingReviewVerifier and StatisticalIntervalAuditor",
      "category_reason": "Tested Python 3.10+ standard-library package pinned by path and SHA-256.",
      "value": {
        "name": "seeded defect and proof carrying review",
        "code": {
          "$code": "seeded-defect-and-proof-carrying-review"
        },
        "runtime": "Python 3.10 or later, standard library only.",
        "commands": [
          "python3 -m proof_carrying_review --bundle BUNDLE.json"
        ],
        "does": [
          "Injects 6 deterministic defect classes into clean candidate bundles to form twin calibration suites.",
          "Enforces proof-carrying review decisions so unfounded refusals and uncalibrated rubber-stamp approvals fail closed.",
          "Computes exact Wilson score intervals and seeded paired bootstrap confidence intervals under ZeroHardcodedValuesAstGate."
        ],
        "does_not": [
          "Perform network I/O, invoke external language models or embed inline magic numbers."
        ],
        "code_license": "MIT"
      },
      "relations": [
        [
          "verifies",
          "spec.seeded-defect-and-review-proof-grid"
        ]
      ]
    },
    {
      "slug": "template.proof-carrying-review-receipt",
      "category": "noun.scene_recipe",
      "role": "template",
      "title": "Template: seeded-defect calibration, proof-carrying review gate and Wilson/bootstrap interval receipt",
      "value": {
        "describes": "a deterministic receipt of 6-class seeded mutant calibration, proof-carrying refusal/approval enforcement, Wilson score intervals and paired bootstrap confidence bounds",
        "schema": {
          "type": "object",
          "additionalProperties": false,
          "required": [
            "defect_class_count",
            "generated_mutant_count",
            "strong_reviewer_status",
            "strong_mutant_kill_rate",
            "strong_clean_fpr",
            "strong_wilson_ci_low",
            "strong_wilson_ci_high",
            "rubber_stamp_status",
            "unfounded_refusal_accepted",
            "rubber_stamp_approval_accepted",
            "proof_carrying_approval_accepted",
            "proof_carrying_refusal_accepted",
            "paired_bootstrap_mean_delta",
            "paired_bootstrap_ci_low",
            "paired_bootstrap_ci_high",
            "paired_statistically_significant"
          ],
          "properties": {
            "defect_class_count": {
              "type": "integer",
              "minimum": 6
            },
            "generated_mutant_count": {
              "type": "integer",
              "minimum": 6
            },
            "strong_reviewer_status": {
              "type": "string",
              "minLength": 1
            },
            "strong_mutant_kill_rate": {
              "type": "number",
              "minimum": 0.0,
              "maximum": 1.0
            },
            "strong_clean_fpr": {
              "type": "number",
              "minimum": 0.0,
              "maximum": 1.0
            },
            "strong_wilson_ci_low": {
              "type": "number",
              "minimum": 0.0,
              "maximum": 1.0
            },
            "strong_wilson_ci_high": {
              "type": "number",
              "minimum": 0.0,
              "maximum": 1.0
            },
            "rubber_stamp_status": {
              "type": "string",
              "minLength": 1
            },
            "unfounded_refusal_accepted": {
              "type": "boolean"
            },
            "rubber_stamp_approval_accepted": {
              "type": "boolean"
            },
            "proof_carrying_approval_accepted": {
              "type": "boolean"
            },
            "proof_carrying_refusal_accepted": {
              "type": "boolean"
            },
            "paired_bootstrap_mean_delta": {
              "type": "number"
            },
            "paired_bootstrap_ci_low": {
              "type": "number"
            },
            "paired_bootstrap_ci_high": {
              "type": "number"
            },
            "paired_statistically_significant": {
              "type": "boolean"
            }
          }
        },
        "example": {
          "defect_class_count": 6,
          "generated_mutant_count": 12,
          "strong_reviewer_status": "trusted_reviewer",
          "strong_mutant_kill_rate": 1.0,
          "strong_clean_fpr": 0.0,
          "strong_wilson_ci_low": 0.7575,
          "strong_wilson_ci_high": 1.0,
          "rubber_stamp_status": "under_sensitive_reviewer",
          "unfounded_refusal_accepted": false,
          "rubber_stamp_approval_accepted": false,
          "proof_carrying_approval_accepted": true,
          "proof_carrying_refusal_accepted": true,
          "paired_bootstrap_mean_delta": 0.22,
          "paired_bootstrap_ci_low": 0.215,
          "paired_bootstrap_ci_high": 0.2263,
          "paired_statistically_significant": true
        }
      }
    },
    {
      "slug": "example.proof-carrying-review-bundle",
      "category": "noun.eval_run",
      "role": "example",
      "synthetic": true,
      "title": "Worked example input: two clean candidate bundles and paired evaluation score vectors",
      "category_reason": "Synthetic fixture recomputed by test_proof_carrying_review.py.",
      "value": {
        "regime": "release_admission_gate",
        "clean_bundles": [
          {
            "bundle_id": "cand-01",
            "artifact": {
              "payload": {
                "record": "cand-01",
                "verified": true
              },
              "declared_sha256": "8ac8f677f97d085b21dcde25b003916674264225b956a3adb87ac034555d129c"
            },
            "boundary": {
              "observed": 6,
              "max_allowed": 8
            },
            "code_snippet": "def aggregate_counts(values: list[int]) -> int:\n    return sum(values)\n",
            "capability": {
              "requested_effect": "write-internal",
              "granted_ceiling": "write-internal"
            },
            "attestation": {
              "drafter_principal": "worker:drafter-cand-01",
              "drafter_key_id": "key_drafter_cand-01",
              "reviewer_principal": "worker:reviewer-cand-01",
              "reviewer_key_id": "key_reviewer_cand-01"
            },
            "sources": {
              "src-1": "Every review refusal requires a reproducible failing check identifier and witness string."
            },
            "citations": [
              {
                "source_id": "src-1",
                "exact_quote": "reproducible failing check identifier and witness string"
              }
            ]
          },
          {
            "bundle_id": "cand-02",
            "artifact": {
              "payload": {
                "record": "cand-02",
                "verified": true
              },
              "declared_sha256": "fa4b653d466a56dc359453244cd8448ecbf54cbd88548382b4df9b435bd3cc7f"
            },
            "boundary": {
              "observed": 8,
              "max_allowed": 8
            },
            "code_snippet": "def aggregate_counts(values: list[int]) -> int:\n    return sum(values)\n",
            "capability": {
              "requested_effect": "write-internal",
              "granted_ceiling": "write-internal"
            },
            "attestation": {
              "drafter_principal": "worker:drafter-cand-02",
              "drafter_key_id": "key_drafter_cand-02",
              "reviewer_principal": "worker:reviewer-cand-02",
              "reviewer_key_id": "key_reviewer_cand-02"
            },
            "sources": {
              "src-1": "Every review refusal requires a reproducible failing check identifier and witness string."
            },
            "citations": [
              {
                "source_id": "src-1",
                "exact_quote": "reproducible failing check identifier and witness string"
              }
            ]
          }
        ],
        "baseline_system_scores": [
          0.62,
          0.65,
          0.6,
          0.64,
          0.61,
          0.63,
          0.59,
          0.64
        ],
        "candidate_system_scores": [
          0.84,
          0.86,
          0.82,
          0.85,
          0.83,
          0.87,
          0.81,
          0.86
        ],
        "bootstrap_seed": 20261031,
        "shows": "Synthetic clean candidate bundles for 6-class mutant expansion and paired system score vectors.",
        "source": "Synthetic. Written for this pack."
      }
    },
    {
      "slug": "example.proof-carrying-review-receipt",
      "category": "noun.eval_run",
      "role": "example",
      "synthetic": true,
      "title": "Worked example output: recomputed 12-mutant kill calibration, proof-carrying gate verdicts and confidence intervals",
      "category_reason": "Synthetic receipt verified by test_proof_carrying_review.py.",
      "value": {
        "conforms_to": {
          "$pin": "template.proof-carrying-review-receipt"
        },
        "instance": {
          "defect_class_count": 6,
          "generated_mutant_count": 12,
          "strong_reviewer_status": "trusted_reviewer",
          "strong_mutant_kill_rate": 1.0,
          "strong_clean_fpr": 0.0,
          "strong_wilson_ci_low": 0.7575,
          "strong_wilson_ci_high": 1.0,
          "rubber_stamp_status": "under_sensitive_reviewer",
          "unfounded_refusal_accepted": false,
          "rubber_stamp_approval_accepted": false,
          "proof_carrying_approval_accepted": true,
          "proof_carrying_refusal_accepted": true,
          "paired_bootstrap_mean_delta": 0.22,
          "paired_bootstrap_ci_low": 0.215,
          "paired_bootstrap_ci_high": 0.2263,
          "paired_statistically_significant": true
        },
        "shows": "Recomputes 12/12 mutant kills (Wilson 95% CI [0.7575, 1.0]), rejects unfounded refusals and rubber-stamp approvals and confirms paired bootstrap gain [0.215, 0.2263].",
        "source": "Synthetic. Recomputed directly from example.proof-carrying-review-bundle."
      },
      "relations": [
        [
          "derives_from",
          "example.proof-carrying-review-bundle"
        ],
        [
          "verifies",
          "template.proof-carrying-review-receipt"
        ]
      ]
    }
  ],
  "verbs": [
    {
      "slug": "generate-seeded-defect-mutant-twins",
      "kind": "verb.generate_seed",
      "title": "Generate deterministic 6-class seeded-defect mutant twins from clean candidate bundles",
      "claimant": "agent",
      "model_use": "none",
      "summary": "Inject digest mismatch, boundary off-by-one, silent exception swallow, capability-tier violation, self-approval principal reuse and ungrounded citation span defects into clean candidate bundles.",
      "outcome": "Deterministic twin calibration suite pairing clean candidate bundles with six targeted defect mutants each.",
      "ready_outcome": "mutant_twins_generated",
      "pinned": [
        "spec.seeded-defect-and-review-proof-grid",
        "rules.mutant-calibration-and-interval-audit",
        "tool.proof-carrying-review"
      ],
      "slots": [
        {
          "name": "clean_candidate_bundle",
          "category": "noun.eval_run",
          "description": "Pin of clean candidate bundles for deterministic mutant synthesis."
        }
      ],
      "produces": [
        {
          "name": "mutant_calibration_receipt",
          "template": "template.proof-carrying-review-receipt"
        }
      ],
      "steps": [
        "Load mutant perturbation offsets and check definitions from default_grid_spec.json.",
        "Apply all six deterministic defect operators to each clean bundle and confirm each mutant trips its exact target check."
      ],
      "acceptance": [
        {
          "id": "six-defect-classes-isolated",
          "check": "Every generated mutant fails its exact expected check_id while leaving unrelated checks passing.",
          "method": "deterministic"
        }
      ]
    },
    {
      "slug": "verify-proof-carrying-review-verdicts",
      "kind": "verb.verify_gate",
      "title": "Verify proof-carrying review refusals and mutant-calibrated approvals",
      "claimant": "agent",
      "model_use": "none",
      "summary": "Enforce no refusal without a reproducible failing check and witness, and no approval unless reviewer mutant kill rate meets the grid-configured threshold.",
      "outcome": "Accepted or rejected review verdict backed by reproduced failing check witnesses and reviewer calibration.",
      "ready_outcome": "review_verdicts_verified",
      "pinned": [
        "spec.seeded-defect-and-review-proof-grid",
        "rules.mutant-calibration-and-interval-audit",
        "guardrail.no-refusal-without-reproducible-witness",
        "guardrail.no-approval-without-mutant-kill-calibration",
        "tool.proof-carrying-review"
      ],
      "slots": [
        {
          "name": "review_decision_bundle",
          "category": "noun.eval_run",
          "description": "Pin of candidate artifacts, reviewer verdicts and calibration twin runs."
        }
      ],
      "produces": [
        {
          "name": "proof_carrying_verdict_receipt",
          "template": "template.proof-carrying-review-receipt"
        }
      ],
      "steps": [
        "Measure reviewer mutant kill rate and clean false-positive rate across the twin suite.",
        "Re-execute cited check IDs on refusals and verify reviewer calibration thresholds on approvals."
      ],
      "acceptance": [
        {
          "id": "unfounded-refusal-and-rubber-stamp-blocked",
          "check": "Refusals without a reproduced failing check witness and approvals from uncalibrated rubber-stamp reviewers both return accepted == false.",
          "method": "deterministic"
        }
      ],
      "follows": [
        "seeded-defect-and-proof-carrying-review.generate-seeded-defect-mutant-twins"
      ]
    },
    {
      "slug": "audit-wilson-and-paired-bootstrap-intervals",
      "kind": "verb.benchmark_harness",
      "title": "Independently audit Wilson score intervals, paired bootstrap deltas and AST magic-literal gates",
      "claimant": "agent",
      "model_use": "none",
      "summary": "Recompute exact Wilson score pass/kill intervals, deterministic paired bootstrap confidence bounds and ZeroHardcodedValuesAstGate conformance under a distinct reviewer.",
      "outcome": "Independent statistical interval and AST conformance verification receipt.",
      "ready_outcome": "intervals_and_ast_verified",
      "review": "distinct_reviewer",
      "pinned": [
        "spec.seeded-defect-and-review-proof-grid",
        "guardrail.no-refusal-without-reproducible-witness",
        "guardrail.no-approval-without-mutant-kill-calibration",
        "tool.proof-carrying-review",
        "template.proof-carrying-review-receipt"
      ],
      "slots": [
        {
          "name": "interval_eval_bundle",
          "category": "noun.eval_run",
          "description": "Pin of twin calibration outcomes and paired baseline-vs-candidate score vectors."
        }
      ],
      "produces": [
        {
          "name": "statistical_audit_receipt",
          "template": "template.proof-carrying-review-receipt"
        }
      ],
      "steps": [
        "Recompute Wilson score confidence bounds and seeded paired bootstrap intervals on the evaluation fixture.",
        "Run scan_directory_for_hardcoded_values across proof_carrying_review and confirm zero violations."
      ],
      "acceptance": [
        {
          "id": "intervals-and-ast-reproduced",
          "check": "scan_directory_for_hardcoded_values returns [] and run_bundle reproduces the exact statistical receipt.",
          "method": "deterministic"
        }
      ],
      "follows": [
        "seeded-defect-and-proof-carrying-review.verify-proof-carrying-review-verdicts"
      ]
    }
  ]
}