PROTOCOL: ADR-VALIDATE (EXECUTION LEVEL 3)

<state_machine> CURRENT_STEP -> REQUIRED_TOOL -> REQUIRED_OUTPUT -> MANDATORY_HALT Step 1: Read Target -> read_file (target ADR + docs/adr/README.md) -> Target ADR content & index state -> Continue to Step 2 Step 2: Audit Rubric -> read_file(templates/TEMPLATE-ADR_VALIDATOR.md) -> 7-criteria weighted rubric evaluation -> Continue to Step 3 Step 3: Score Report -> format using TEMPLATE-ADR_VALIDATOR.md -> Scored validation report (High / Medium / Low) -> 🛑 HALT TURN (Present audit) </state_machine>

<execution_invariants>

1. WORKFLOW IMMUTABILITY: Execute all 3 steps and all 7 rubric criteria in strict order (workflow structure = immutable | step reasoning = flexible).
2. TEMPLATE LOCK: You MUST call read_file on templates/TEMPLATE-ADR_VALIDATOR.md before producing the validation report and preserve its exact rubric sections and weighted scoring structure.
3. ESCAPE HATCH: If the situation does not fit the contract, do not silently improvise and do not force-fit: declare the deviation in one line and ask. </execution_invariants>

<output_contract>

• SOURCE_TEMPLATE: templates/TEMPLATE-ADR_VALIDATOR.md (MUST call read_file before producing report)
• SCHEMA_INVARIANCE: Preserve all 7 validation criteria sections, the weighted Quality Score Summary table (3.0 scale), and the Validation Summary verbatim.
• ESCAPE_HATCH: Never silently omit or invent rubric sections. If a section (such as Batch Validation Summary when auditing a single ADR) does not apply, preserve the heading and write "N/A — Single ADR audit", or omit only sections explicitly marked (Multiple ADRs). </output_contract>

📋 Step-by-Step Workflow

1. Identify Target ADR File

Locate the file in docs/adr/ specified by the user (or prompt for the path if not provided).

2. Audit Against the 7-Criteria Weighted Quality Rubric (templates/TEMPLATE-ADR_VALIDATOR.md)

Criteria 1: Metadata & Formal Compliance (Weight: 10%)

• Status field present and valid (Proposed, Accepted, Approved, Deprecated, Superseded).
• Date field present (YYYY-MM-DD format).
• Decision makers identified.
• Supersedes / Amended By (or Amends) fields completed with two-way links where applicable.
• ADR follows standard template structure and 3-digit sequential numbering (0XX).

Criteria 2: Context & Problem Statement (Weight: 20%)

• Context is clear and unbiased (value-neutral).
• Problem statement is specific and well-defined.
• References specific pain points or constraints.
• Decision drivers are explicitly stated.
• Links back to inception artifacts or business requirements.

Criteria 3: Options & Analysis (Weight: 20%)

• At least 2–3 viable alternatives considered.
• No strawman options (all options are realistic).
• Each option has pros and cons documented (Good, because... / Bad, because...).
• Options are compared against decision drivers and trade-offs are explicitly discussed.

Criteria 4: Decision Outcome (Weight: 20%)

• Chosen option is clearly identified.
• Justification is logical, evidence-based, and references decision drivers.
• Decision aligns with project constraints (cost, timeline, architecture boundaries).

Criteria 5: Consequences Documentation (Weight: 10%)

• Positive consequences explicitly listed.
• Negative consequences explicitly listed.
• Risks identified and documented with actionable mitigations.

Criteria 6: Traceability & References (Weight: 10%)

• Links to inception artifacts, related ADRs, and external documentation/benchmarks.
• Record is listed in docs/adr/README.md with working relative links.

Criteria 7: AI & Agentic Ergonomics (AX) [Mandatory] (Weight: 10%)

• LLM Corpus Alignment states an estimated training-corpus match (High >80% / Medium 30-80% / Atypical/Exotic <30%).
• Expected Agent Bias | Automated Guardrail table pairs each expected agent deviation with a named automated countermeasure.
• Rows marked — are explicitly justified as accepted manual-review risks.
• Supervision & Guardrail Tax (Low, Medium, High) is assessed.

3. Assign Quality Score & Provide Feedback

Score each of the 7 criteria (High = 3, Medium = 2, Low = 1), compute the weighted score out of 3.0 ([X]/7 criteria at High level), and assign the rating:

• High Quality (Accept): 6–7 criteria at High level, no Low ratings.
• Medium Quality (Accept with Revisions): 4–5 criteria at High level, or 1–2 Medium ratings. Provide concrete action items.
• Low Quality (Reject / Revise): More than 2 criteria at Medium or any Low rating (e.g., missing AX section or strawman options). Block adoption until remediated.

<self_validation_gate> Before ending your turn, verify every item below with concrete evidence (not just a checkmark). If any item cannot be cited with real evidence from this turn, DO NOT produce your final response — fix the violation first:

• Target Read Evidence: Exact target ADR path(s) and docs/adr/README.md read via read_file: <paths>
• Template Read Evidence: Exact path passed to read_file (templates/TEMPLATE-ADR_VALIDATOR.md): <path>
• Rubric Score Evidence: All 7 weighted criteria scored ([X]/7 at High level, weighted total [Y]/3.0): <scores> </self_validation_gate>