mm_verify

Run one phase of the iterative verify-spec pipeline — start_run/finish_run delimit a measured run; the other phases perform checks, classification, persistence, confidence, and promotion. phase must be one of the enum values below; see the verify-* sources under src/tools/ for per-phase argument shapes.

At a Glance

• Category: Core
• Source: Built-in
• Availability: Listed by default
• Required parameters: 1
• Total parameters: 68

Parameters

Parameter	Type	Required	Description
phase	string (enum: check_triage, start_run, finish_run, prepare_evidence, classify_findings, compute_confidence, convergence_step, persist_findings, write_summary, run_checks, parse_requirements, evaluate, classify_risk, record_outcome, promote, light_check)	Yes	Pipeline phase to invoke. Each phase carries its own per-call arguments — see the verify-* handler sources.
project_path	string	No	check_triage: absolute path to the project root (the directory containing .mm/).
spec_folder	string	No	check_triage: spec folder name (spec_path accepted as alias).
verification_run_id	string	No	Stable id shared by every phase and captured turn in one verification run.
spec_slug	string	No	start_run/finish_run: canonical spec folder slug.
workflow	string	No	Telemetry label; verify-spec for this pipeline.
base_sha	string | null	No
head_sha	string | null	No
iteration	integer	No
agent_role	string	No
parent_session_id	string	No
request_id	string	No
status	string (enum: verified, verified-with-caveats, gaps-remain)	No	finish_run: final verification verdict.
measurement_complete	boolean	No
measurement_reason	string	No
findings	array of object	No	classify_findings: verifier-emitted Finding records (id, kind, spec_section_ref, code_location, failure_mode, reproduction, evidence, tier) — see verify-classify-findings.ts.
runs_completed	integer	No
t1_count	integer	No
t2_count	integer	No
t3_count	integer	No
t4_count	integer	No
adversarial_null_count	integer	No
adversarial_malformed	integer	No
decision	string	No	compute_confidence: continue | converged | did_not_converge.
consecutive_stable	integer	No
run_index	integer	No
current_t12_ids	array of string	No
prior_t12_ids	array of string	No
floor	integer	No
ceiling	integer	No
run_label	string	No	persist_findings: human-meaningful run label recorded on new spec-gap / backlog rows (audit-only).
now	string	No	persist_findings: ISO-8601 timestamp stamped on new rows. Defaults to now; injectable for tests.
run_timestamp	string	No	write_summary: ISO-8601 timestamp for verify-summary.json. Defaults to now.
tier_counts	object	No	write_summary: { t1, t2, t3, t4 } counts of classified findings this run.
spec_gap_count	integer	No
confidence	object	No	write_summary: { bucket, numeric, rationale } from the compute_confidence phase.
open_backlog_count	integer	No
open_spec_gaps_count	integer	No
code_path	string	No	run_checks: absolute path to the checkout/worktree holding the code under verification. Defaults to project_path; the checks run here.
verifier_model	string	No	evaluate: model name recorded as verified_by.
scorecard	array of object	No	evaluate: one row per R<n> — { id, status: pass|partial|miss|regression|pass-pre-existing, citation?, note? }.
checks	object	No	evaluate: run_checks output (or the subset: passed_ids/failed_ids/skipped_ids/env_skipped_ids/all_skipped_env).
adversarial	array of object	No	evaluate: adversarial-prompt rows — { prompt, outcome: finding|null, spec_section_ref?, justification?, finding_id? }.
open_backlog	array of object	No	evaluate: open check_triage entries — { cluster_id_short, failure_mode, spec_section_ref?, status?, actor? }.
diff	object	No	evaluate: { files_changed, files_mapped, unmapped }.
strict	boolean	No	evaluate: 0-tolerance for partially_supported claims when true.
mode	string	No	evaluate: recorded on summary.json.
risk_tier	string (enum: low, medium, high)	No	evaluate: recorded on summary.json.
risk_reasons	array of string	No
touched_paths	array of string	No	classify_risk: repo-relative paths touched by the diff under verification.
total_changed_lines	integer	No
touched_requirement_ids	array of string	No
critical_requirement_ids	array of string	No
floor_confidence	string (enum: low, medium, high)	No	classify_risk: floor-pass confidence, when the floor pass has already run.
floor_flagged_high_risk	boolean	No
max_total_lines	integer	No
max_files	integer	No
extra_high_risk_patterns	array of string	No
style_only_paths	array of string	No	classify_risk: touched paths whose change is style-only; all touched paths must be listed for the trivial tier.
max_trivial_lines	integer	No
max_trivial_files	integer	No
light_checks	array of object	No	light_check: the mechanical checks the caller ran — { name, ok, detail? }.
outcome	string (enum: spec-verified, pr-merged)	No	record_outcome: which value boundary this row records.
ref	string	No	record_outcome: what the outcome refers to — spec folder, or PR number/URL.
session_ids	array of string	No
started_at	string	No
ended_at	string	No
event_id	string	No

Example

{
  "tool": "mm_verify",
  "arguments": {
    "phase": "check_triage",
    "project_path": "/Users/you/projects/my-app",
    "spec_folder": "example-spec_folder"
  }
}

Related Tools

• mm_search - Full-text search across all entity types (specs, features, issues, notes, ide…
• mm_configure_integration - Configure an external service integration (github, trello, grain, gmail, noti…
• mm_sync_notion - Bidirectional sync between a Notion database and .mm/specs/ or .mm/notes/. Pu…
• mm_daily_brief - Generate a structured morning summary with active tasks, blockers, recently c…
• mm_revert_spec_completion - Revert a spec’s completion status in its parent feature. Tasks stay [x] but t…