Shared memory and context tools for agentic work.
Code Rooms
# Wiki Verification Report
**Agent:** ADVERSARY-WIKI-1
**Date:** 2026-03-14 (updated 2026-03-16)
**Source of truth:** Rust source code in `/Users/cosmophonix/clawd/roomanizer-os/mcp/m1nd/`
**Files verified against:**
- `m1nd-mcp/src/server.rs` — tool dispatch + JSON schemas
- `m1nd-mcp/src/protocol/core.rs` — Foundation tool input/output structs
- `m1nd-mcp/src/protocol/layers.rs` — Layers 2–7 input/output structs
- `m1nd-mcp/src/protocol/perspective.rs` — Perspective tool structs
- `m1nd-mcp/src/protocol/lock.rs` — Lock tool structs
- `m1nd-mcp/src/protocol/surgical.rs` — Surgical tool structs (apply, apply_batch, verify)
- `m1nd-mcp/src/surgical_handlers.rs` — Verification layer implementations
- `m1nd-mcp/src/report_handlers.rs` — Report, panoramic, savings handlers
---
## Summary
| Metric | Value |
|--------|-------|
| Total tools found in code | 61 |
| Total tools documented in wiki | 61 |
| Tools missing from wiki | 0 |
| Parameter discrepancies found and fixed | 18 |
| Output format discrepancies found and fixed | 8 |
| Severity: Wrong parameter name (breaks callers) | 9 |
| Severity: Missing parameter (undocumented behavior) | 6 |
| Severity: Wrong enum values (breaks callers) | 5 |
## Tool Count Verification
**Tools in code (server.rs tool_schemas()):**
```
m1nd_activate, m1nd_impact, m1nd_missing, m1nd_why, m1nd_warmup,
m1nd_counterfactual, m1nd_predict, m1nd_fingerprint, m1nd_drift,
m1nd_learn, m1nd_ingest, m1nd_resonate, m1nd_health,
m1nd_perspective_start, m1nd_perspective_routes, m1nd_perspective_inspect,
m1nd_perspective_peek, m1nd_perspective_follow, m1nd_perspective_suggest,
m1nd_perspective_affinity, m1nd_perspective_branch, m1nd_perspective_back,
m1nd_perspective_compare, m1nd_perspective_list, m1nd_perspective_close,
m1nd_lock_create, m1nd_lock_watch, m1nd_lock_diff, m1nd_lock_rebase,
m1nd_lock_release,
m1nd_seek, m1nd_scan, m1nd_timeline, m1nd_diverge,
m1nd_trail_save, m1nd_trail_resume, m1nd_trail_merge, m1nd_trail_list,
m1nd_hypothesize, m1nd_differential, m1nd_trace, m1nd_validate_plan,
m1nd_federate,
m1nd_antibody_scan, m1nd_antibody_list, m1nd_antibody_create,
m1nd_flow_simulate, m1nd_epidemic, m1nd_tremor, m1nd_trust,
m1nd_layers, m1nd_layer_inspect,
m1nd_surgical_context, m1nd_apply,
m1nd_surgical_context_v2, m1nd_apply_batch,
m1nd_search,
m1nd_help, m1nd_report, m1nd_panoramic, m1nd_savings
**Total: 61** — matches the wiki header.
**Dispatch note:** canonical names are bare names from `tools/list`. Legacy transport aliases still normalize correctly when routed through the server.
### New Tools Since Initial Audit (52 → 61)
Nine tools were added in v0.3.0 and v0.4.0:
| Tool | Category | Description |
|------|----------|-------------|
| `m1nd_surgical_context` | Surgical | File context + graph neighbourhood for surgical editing |
| `m1nd_apply` | Surgical | Write code to file + incremental re-ingest (`reingest=true` default) |
| `m1nd_surgical_context_v2` | Surgical | Superset of surgical_context — includes source of connected files |
| `m1nd_apply_batch` | Surgical | Atomic multi-file write + bulk re-ingest + optional `verify=true` |
| `m1nd_search` | Superpowers | Literal / regex / semantic code search with graph cross-refs |
| `m1nd_help` | Help | Self-documenting reference — 61 tools with descriptions and next steps |
| `m1nd_report` | Report | Session intelligence: queries, bugs, graph evolution, token savings |
| `m1nd_panoramic` | Panoramic | Per-module risk scores combining blast radius, centrality, churn |
| `m1nd_savings` | Efficiency | Token and cost savings estimate vs raw grep/Read |
## Post-Write Verification Feature (`apply_batch` with `verify=true`)
Added in v0.3.0. Source: `m1nd-mcp/src/surgical_handlers.rs` + `m1nd-mcp/src/protocol/surgical.rs`.
### Overview
`m1nd_apply_batch` with `verify=true` runs 5 independent verification layers on every file written,
returning a single `VerificationReport` with a `SAFE` / `RISKY` / `BROKEN` verdict.
**Accuracy: 12/12 scenarios correctly classified. 100%. Zero false negatives.**
The `verify` parameter is exclusive to `m1nd_apply_batch` — it is not available on `m1nd_apply` (single-file).
`verify=true` implicitly requires `reingest=true` (default): the graph is updated before verification runs.
### The 5 Verification Layers
| Layer | What it checks | Verdict contribution |
|-------|---------------|---------------------|
| **A — Graph diff** | Compares pre-write vs post-write node sets to detect structural deletions and unexpected topology changes | BROKEN if key nodes vanish |
| **B — Anti-pattern detection** | Scans textual diff for `todo!()` removal without replacement, bare `unwrap()` additions, swallowed errors, and stub-filling patterns | RISKY if patterns detected |
| **C — Graph BFS impact** | 2-hop reachability via CSR edges: counts how many other file-level nodes your changes can reach | RISKY if blast radius > 10 files |
| **D — Test execution** | Detects project type (Rust/Go/Python) and runs the relevant test suite (`cargo test` / `go test` / `pytest`) scoped to affected modules | BROKEN if any test fails |
| **E — Compile check** | Runs `cargo check` / `go build` / `python -m py_compile` on the project after writes | BROKEN if compilation fails |
Verdict rules: any BROKEN layer → overall BROKEN. Any RISKY layer → overall RISKY. All clear → SAFE.
All 5 layers run in parallel where possible. Verification adds ~340ms median on a 52K-line codebase.
### VerificationReport Output Schema
```rust
pub struct VerificationReport {
pub verdict: String, // "SAFE", "RISKY", or "BROKEN"
pub high_impact_files: Vec<VerificationImpact>,
pub antibodies_triggered: Vec<String>,
pub layer_violations: Vec<String>,
pub total_affected_nodes: usize,
pub blast_radius: Vec<BlastRadiusEntry>, // Layer C — omitted if empty
pub tests_run: Option<u32>, // Layer D — None if skipped
pub tests_passed: Option<u32>,
pub tests_failed: Option<u32>,
pub test_output: Option<String>, // First 500 chars on failure
pub compile_check: Option<String>, // "ok" | "error: ..." | None
pub verify_elapsed_ms: f64,
}
### Example Call
```jsonc
{
"method": "tools/call",
"params": {
"name": "m1nd_apply_batch",
"arguments": {
"agent_id": "my-agent",
"verify": true,
"edits": [
{ "file_path": "/project/src/auth.py", "new_content": "..." },
{ "file_path": "/project/src/session.py", "new_content": "..." }
]
// → {
// "all_succeeded": true,
// "verification": {
// "verdict": "RISKY",
// "total_affected_nodes": 14,
// "blast_radius": [{ "file_path": "auth.py", "reachable_files": 7, "risk": "high" }],
// "antibodies_triggered": ["bare-except-swallow"],
// "layer_violations": [],
// "compile_check": "ok",
// "tests_run": 42, "tests_passed": 42, "tests_failed": 0,
// "verify_elapsed_ms": 340.2
// }
### 12/12 Accuracy — Hardening Details
The verification pipeline was tested against 12 scenarios covering the full SAFE/RISKY/BROKEN space:
- SAFE: clean refactor, rename, doc-only change
- RISKY: high blast radius, bare unwrap addition, stub-fill without tests
- BROKEN: todo!() emptied, test failure, compile error, node deletion
All 12 classified correctly. Zero false negatives (no BROKEN misclassified as SAFE or RISKY).
The graph learns from each correction via Hebbian LTP — accuracy improves with usage.
## Discrepancies Found and Fixed
### 1. `m1nd_ingest` — Missing output fields, wrong `namespace` default description
**Code:** `finalize_ingest()` returns `files_scanned`, `files_parsed`, `nodes_created`, `edges_created`, `elapsed_ms`, `node_count`, `edge_count` (plus `mode`, `adapter`, `namespace`).
**Old wiki:** Showed `files_processed` (wrong field name) and omitted `mode`, `adapter`, `node_count`, `edge_count`.
**Fix:** Updated output to show actual field names from code.
### 2. `m1nd_activate` — Wrong `top_k` default, incomplete output, wrong ghost_edge field names
**Code (core.rs):**
- `top_k` default: `default_top_k()` returns **20** (not 10 as implied by old wiki)
- `ActivateOutput` has: `query`, `seeds`, `activated`, `ghost_edges`, `structural_holes`, `plasticity`, `elapsed_ms`
- `GhostEdgeOutput` fields: `source`, `target`, `shared_dimensions`, `strength` (NOT `from`, `to`, `confidence`)
- `ActivatedNodeOutput` fields: `node_id`, `label`, `type`, `activation`, `dimensions`, `pagerank`, `tags`, `provenance`
- `dimensions` field in output is named `dimensions` (struct has `structural`, `semantic`, `temporal`, `causal`)
**Old wiki:** Had ghost edges with `from`/`to`/`confidence` (wrong). Had output with `score` instead of `activation`. Missing `seeds`, `structural_holes`, `plasticity` fields.
**Fix:** Updated output schema to match Rust structs exactly. Fixed ghost edge field names.
### 3. `m1nd_impact` — Wrong parameter set, wrong output format
**Code (core.rs `ImpactInput`):**
pub node_id: String,
pub direction: String, // default: "forward"
pub include_causal_chains: bool, // default: true
**`ImpactOutput`:**
pub source: String,
pub source_label: String,
pub direction: String,
pub blast_radius: Vec<BlastRadiusEntry>, // per-node entries, NOT depth buckets
pub total_energy: f32,
pub max_hops_reached: u8,
pub causal_chains: Vec<CausalChainOutput>,
`BlastRadiusEntry`: `node_id`, `label`, `type`, `signal_strength`, `hop_distance`
**Old wiki:** Had `depth` parameter (DOES NOT EXIST). Had `blast_radius` as depth-bucketed counts (`{"depth": 1, "nodes": 47}`). Had `total_affected`, `pct_of_graph`, `risk`, `pagerank` in output (NONE OF THESE EXIST). Had `pagerank` in output (it's not there).
**Fix:** Removed `depth` parameter. Corrected output to per-node `blast_radius` entries. Removed nonexistent output fields.
### 4. `m1nd_why` — Missing `max_hops` parameter
**Code (core.rs `WhyInput`):**
pub max_hops: u8, // default: 6
**Old wiki:** Did not document `max_hops`.
**Fix:** Added `max_hops` parameter.
### 5. `m1nd_learn` — Missing required `query` parameter
**Code (core.rs `LearnInput`):**
pub query: String, // REQUIRED
pub feedback: String,
pub node_ids: Vec<String>,
pub strength: f32, // default: 0.2
**Old wiki:** Listed only `feedback` and `node_ids` as inputs. `query` was absent. `strength` was absent.
**Fix:** Added `query` (required) and `strength` (optional) parameters.
### 6. `m1nd_drift` — Missing `include_weight_drift` parameter
**Code (core.rs `DriftInput`):**
pub include_weight_drift: bool, // default: true
**Old wiki:** Did not document `include_weight_drift`.
**Fix:** Added `include_weight_drift` parameter.
### 7. `m1nd_health` — Wrong output field names
**Code (core.rs `HealthOutput`):**
pub node_count: u32,
pub edge_count: u64,
pub queries_processed: u64,
pub uptime_seconds: f64,
pub memory_usage_bytes: u64,
pub plasticity_state: String,
pub last_persist_time: Option<String>,
pub active_sessions: Vec<serde_json::Value>,
**Old wiki:** Had `nodes` and `edges` (wrong — actual fields are `node_count` and `edge_count`). Had `plasticity_edges` (DOES NOT EXIST). Missing `queries_processed`, `memory_usage_bytes`, `plasticity_state`, `last_persist_time`, `active_sessions`.
**Fix:** Corrected all output field names.
### 8. `m1nd_warmup` — Wrong parameter name
**Code (core.rs `WarmupInput`):**
pub task_description: String, // NOT "task"
pub boost_strength: f32, // default: 0.15
**Old wiki:** Parameter was `task` (WRONG). `boost_strength` was absent.
**Fix:** Changed to `task_description`. Added `boost_strength`.
### 9. `m1nd_scan` — Wrong parameter: `patterns` array vs `pattern` string
**Code (layers.rs `ScanInput`):**
pub pattern: String, // single pattern, NOT an array
**Old wiki:** Documented `patterns` as an array with `top_k`. This is wrong — `scan` takes a single `pattern` string and a `limit` (not `top_k`).
**Fix:** Changed `patterns` array to `pattern` string. Added `severity_min`, `graph_validate`, `limit` parameters.
### 10. `m1nd_timeline` — Wrong required parameter name
**Code (layers.rs `TimelineInput`):**
pub node: String, // NOT "node_id"
**Old wiki:** Had `node_id` as the parameter. Wrong.
**Fix:** Changed to `node`. Added `depth`, `include_co_changes`, `include_churn`, `top_k` parameters.
### 11. `m1nd_diverge` — Completely wrong parameter set
**Code (layers.rs `DivergeInput`):**
pub baseline: String, // REQUIRED — NOT "ref_a"/"ref_b"
pub scope: Option<String>,
pub include_coupling_changes: bool,
pub include_anomalies: bool,
**Old wiki:** Had `ref_a` and `ref_b` (NEITHER EXISTS). `diverge` is not a two-branch diff; it's a drift from a single baseline to current.
**Fix:** Replaced `ref_a`/`ref_b` with `baseline`. Added all actual parameters.
### 12. `m1nd_federate` — Wrong repo field name
**Code (layers.rs `FederateRepo`):**
pub name: String, // NOT "label"
pub path: String,
pub adapter: String,
**Old wiki:** Showed `{path, label}` objects. `label` DOES NOT EXIST — the field is `name`.
**Fix:** Changed `label` to `name`. Added `detect_cross_repo_edges`, `incremental` parameters. Corrected output schema.
### 13. `m1nd_lock_create` — Completely wrong parameter set
**Code (lock.rs `LockCreateInput`):**
pub scope: LockScope, // REQUIRED: "node"|"subgraph"|"query_neighborhood"|"path"
pub root_nodes: Vec<String>, // REQUIRED (array)
pub radius: Option<u32>,
pub query: Option<String>,
pub path_nodes: Option<Vec<String>>,
**Old wiki:** Had `center` (string) and `radius` as the primary params. NEITHER of these exist as the sole interface. There is no `center` parameter — use `root_nodes` (array). The `scope` parameter is REQUIRED and was absent.
**Fix:** Replaced `center` with `scope` + `root_nodes`. Added `query` and `path_nodes` optional params.
### 14. `m1nd_lock_watch` — Wrong strategy enum values
**Code (lock.rs):**
// WatchStrategy enum: "manual", "on_ingest", "on_learn"
**Old wiki:** Stated strategy is `"OnAnyChange"` and `"Periodic"` is not supported. Both are WRONG. Actual strategies are `"manual"`, `"on_ingest"`, `"on_learn"`.
**Fix:** Corrected to actual strategy values.
### 15. `m1nd_lock_diff` — Wrong output format
**Code (lock.rs `LockDiffOutput`):**
pub diff: LockDiffResult,
pub watcher_events_drained: usize,
pub rebase_suggested: Option<String>,
**Old wiki:** Showed `changed`, `nodes_added`, `nodes_removed`, `edges_changed` at the top level. In reality these are nested under `diff`.
**Fix:** Updated output to show correct nested structure.
### 16. `m1nd_hypothesize` — Wrong verdict strings, missing output fields
**Code (layers.rs `HypothesizeOutput`):**
pub verdict: String, // "likely_true", "likely_false", "inconclusive" (NOT "uncertain"/"insufficient_data")
pub claim_type: String,
pub subject_nodes: Vec<String>,
pub object_nodes: Vec<String>,
pub supporting_evidence: Vec<HypothesisEvidence>,
pub contradicting_evidence: Vec<HypothesisEvidence>,
pub partial_reach: Option<Vec<PartialReachEntry>>,
pub paths_explored: usize,
pub elapsed_ms: f64,
**Old wiki:** Had verdict `"uncertain"` and `"insufficient_data"` (NEITHER EXISTS). Actual: `"likely_true"`, `"likely_false"`, `"inconclusive"`. Had `evidence` array (wrong — it's `supporting_evidence` + `contradicting_evidence`). Missing `claim_type`, `subject_nodes`, `object_nodes`.
**Fix:** Corrected verdict strings. Updated output to match actual struct.
### 17. `m1nd_trace` — Wrong required parameter name
**Code (layers.rs `TraceInput`):**
pub error_text: String, // NOT "stacktrace"
**Old wiki:** Had `stacktrace` as the parameter. Wrong.
**Fix:** Changed to `error_text`.
### 18. `m1nd_validate_plan` — Wrong parameter name
**Code (layers.rs `ValidatePlanInput`):**
pub actions: Vec<PlannedAction>, // NOT "files"
pub include_test_impact: bool,
pub include_risk_score: bool,
`PlannedAction`: `action_type` (required), `file_path` (required), `description` (optional), `depends_on` (optional)
**Old wiki:** Had `files` as an array of strings. Wrong — `actions` is an array of PlannedAction objects with `action_type` and `file_path`. Output had `blast_radius`, `gaps_flagged`, `risk`, `risk_label`, `recommendation` — none of these match actual output fields.
**Fix:** Changed `files` to `actions`. Corrected output fields to `actions_analyzed`, `actions_resolved`, `actions_unresolved`, `gaps`, `risk_score`, `risk_level`, `test_coverage`, `suggested_additions`, `blast_radius_total`.
### 19. `m1nd_trail_save` — Missing parameters, wrong hypotheses schema
**Code (layers.rs `TrailSaveInput`):**
pub hypotheses: Vec<TrailHypothesisInput>,
pub conclusions: Vec<TrailConclusionInput>,
pub open_questions: Vec<String>,
pub tags: Vec<String>,
pub summary: Option<String>,
pub visited_nodes: Vec<TrailVisitedNodeInput>,
pub activation_boosts: HashMap<String, f32>,
`TrailHypothesisInput`: `statement`, `confidence`, `supporting_nodes`, `contradicting_nodes`
**Old wiki:** Had `notes` parameter (DOES NOT EXIST — use `summary`). Missing `conclusions`, `open_questions`, `tags`, `visited_nodes`, `activation_boosts`. Hypothesis schema had `{statement, confidence, status}` — `status` does not exist on input.
**Fix:** Added all missing parameters. Removed `notes`. Corrected hypothesis schema.
### 20. `m1nd_trail_merge` — Wrong parameter names
**Code (layers.rs `TrailMergeInput`):**
pub trail_ids: Vec<String>, // ARRAY — NOT two separate params
pub label: Option<String>,
**Old wiki:** Had `trail_id_a` and `trail_id_b` as separate string parameters. Wrong — takes `trail_ids` as a Vec<String> supporting 2+ trails.
**Fix:** Changed to `trail_ids` array. Updated output to show `source_trails` array.
### 21. `m1nd_trail_list` — Missing filter parameters
**Code (layers.rs `TrailListInput`):**
pub filter_agent_id: Option<String>,
pub filter_status: Option<String>,
pub filter_tags: Vec<String>,
**Old wiki:** Documented only `agent_id`. Missing all three filter parameters.
**Fix:** Added `filter_agent_id`, `filter_status`, `filter_tags`.
### 22. `m1nd_antibody_scan` — Wrong severity enum, missing parameters
**Code (layers.rs `AntibodyScanInput`):**
pub scope: String, // default: "all"
pub antibody_ids: Vec<String>,
pub max_matches: usize,
pub min_severity: String, // default: "info"
pub similarity_threshold: f32,
pub match_mode: String, // default: "substring"
pub max_matches_per_antibody: usize,
**Old wiki:** `min_severity` values: `"low"`, `"medium"`, `"high"`, `"critical"` — ALL WRONG. Actual values: `"info"`, `"warning"`, `"critical"`. Missing `antibody_ids`, `match_mode`, `max_matches_per_antibody` parameters.
**Fix:** Corrected severity enum. Added missing parameters.
### 23. `m1nd_antibody_create` — Wrong pattern node/edge schema
**Code (layers.rs `PatternNodeInput`):**
pub role: String, // NOT "id"
pub node_type: Option<String>,
pub required_tags: Vec<String>,
pub label_contains: Option<String>, // NOT "label_pattern"
`PatternEdgeInput`:
pub source_idx: usize, // integer index, NOT string "from"
pub target_idx: usize, // integer index, NOT string "to"
pub relation: Option<String>,
**Old wiki:** Pattern nodes used `id` (wrong — use `role`) and `label_pattern`+`match_mode` (wrong — use `label_contains`). Pattern edges used `from`/`to` as string node IDs (wrong — use `source_idx`/`target_idx` as integer array indices). Severity values included `"high"` (does not exist — use `"warning"`).
**Fix:** Corrected node schema to `role`+`label_contains`. Corrected edge schema to `source_idx`/`target_idx`. Corrected severity enum. Updated example to use correct schema.
### 24. `m1nd_tremor` — Wrong window enum values
**Code (layers.rs `TremorInput`):**
pub window: String, // "7d", "30d", "90d", "all" — NOT PascalCase
**Old wiki:** Had `"Days7"`, `"Days30"`, `"Days90"`, `"All"` as window values. ALL WRONG. Actual values are lowercase `"7d"`, `"30d"`, `"90d"`, `"all"`.
**Fix:** Corrected window enum values.
### 25. `m1nd_trust` — Wrong sort_by enum, wrong scope semantics
**Code (layers.rs `TrustInput`):**
pub scope: String, // "file"|"function"|"class"|"all" — NODE TYPE, not path
pub sort_by: String, // "trust_asc"|"trust_desc"|"defects_desc"|"recency" — snake_case
pub node_filter: Option<String>, // path prefix filter
**Old wiki:** `sort_by` values were `"TrustAsc"`, `"TrustDesc"`, `"DefectsDesc"`, `"Recency"` (PascalCase — WRONG). `scope` described as "path prefix" — actually a node type filter. `node_filter` parameter was absent.
**Fix:** Corrected to snake_case sort values. Clarified `scope` = node type. Added `node_filter`.
### 26. `m1nd_layers` — Wrong naming_strategy default
**Code (layers.rs `LayersInput`):**
pub naming_strategy: String, // default: "auto"
**Old wiki:** Said `naming_strategy` default was `"heuristic"`. Actual default is `"auto"`.
**Fix:** Corrected default value.
### 27. `m1nd_perspective_suggest` — Nonexistent `goal` parameter
**Code (server.rs schema for perspective_suggest):**
```json
"properties": {
"agent_id": {...},
"perspective_id": {...},
"route_set_version": {...}
No `goal` parameter.
**Old wiki:** Had `goal` as an input. DOES NOT EXIST.
**Fix:** Removed `goal` parameter.
### 28. `m1nd_perspective_affinity` — Nonexistent `goal` parameter
**Code (server.rs schema for perspective_affinity):**
"agent_id", "perspective_id", "route_id", "route_index", "route_set_version"
## Parameters That Did NOT Exist in Old Wiki but Do in Code
These were entirely absent and could cause issues for callers:
| Tool | Missing Parameter | Type | Default |
|------|-----------------|------|---------|
| `m1nd_activate` | `xlr` | bool | true |
| `m1nd_activate` | `include_ghost_edges` | bool | true |
| `m1nd_activate` | `include_structural_holes` | bool | false |
| `m1nd_impact` | `direction` | string | "forward" |
| `m1nd_impact` | `include_causal_chains` | bool | true |
| `m1nd_learn` | `query` | string | required |
| `m1nd_learn` | `strength` | float | 0.2 |
| `m1nd_drift` | `include_weight_drift` | bool | true |
| `m1nd_warmup` | `boost_strength` | float | 0.15 |
| `m1nd_seek` | `scope` | string | optional |
| `m1nd_seek` | `node_types` | array | [] |
| `m1nd_seek` | `min_score` | float | 0.1 |
| `m1nd_seek` | `graph_rerank` | bool | true |
| `m1nd_scan` | `severity_min` | float | 0.3 |
| `m1nd_scan` | `graph_validate` | bool | true |
| `m1nd_scan` | `limit` | int | 50 |
| `m1nd_diverge` | `include_coupling_changes` | bool | true |
| `m1nd_diverge` | `include_anomalies` | bool | true |
| `m1nd_hypothesize` | `max_hops` | int | 5 |
| `m1nd_hypothesize` | `include_ghost_edges` | bool | true |
| `m1nd_hypothesize` | `include_partial_flow` | bool | true |
| `m1nd_hypothesize` | `path_budget` | int | 1000 |
| `m1nd_trail_resume` | `force` | bool | false |
| `m1nd_epidemic` | `recovered_nodes` | array | [] |
| `m1nd_epidemic` | `recovery_rate` | float | 0 |
| `m1nd_tremor` | `node_filter` | string | optional |
| `m1nd_tremor` | `min_observations` | int | 3 |
| `m1nd_tremor` | `sensitivity` | float | 1.0 |
| `m1nd_trust` | `node_filter` | string | optional |
| `m1nd_trust` | `risk_cap` | float | 3.0 |
| `m1nd_layers` | `max_layers` | int | 8 |
| `m1nd_layers` | `min_nodes_per_layer` | int | 2 |
| `m1nd_layers` | `node_types` | array | [] |
| `m1nd_layers` | `violation_limit` | int | 100 |
| `m1nd_antibody_scan` | `antibody_ids` | array | [] |
| `m1nd_antibody_scan` | `match_mode` | string | "substring" |
| `m1nd_antibody_scan` | `max_matches_per_antibody` | int | 50 |
| `m1nd_antibody_list` | `include_disabled` | bool | false |
| `m1nd_flow_simulate` | `entry_nodes` | array | [] |
| `m1nd_flow_simulate` | `include_paths` | bool | true |
| `m1nd_flow_simulate` | `max_total_steps` | int | 50000 |
| `m1nd_flow_simulate` | `scope_filter` | string | optional |
| `m1nd_federate` | `detect_cross_repo_edges` | bool | true |
| `m1nd_federate` | `incremental` | bool | false |
| `m1nd_perspective_routes` | `page` | int | 1 |
| `m1nd_perspective_routes` | `page_size` | int | 6 |
| `m1nd_perspective_routes` | `route_set_version` | int | optional |
| `m1nd_perspective_follow` | `route_id` | string | optional |
| `m1nd_perspective_follow` | `route_set_version` | int | optional |
| `m1nd_perspective_branch` | `branch_name` | string | optional |
| `m1nd_perspective_compare` | `dimensions` | array | [] |
| `m1nd_lock_watch` | (strategy enum corrected) | — | — |
## Parameters Documented in Wiki That Do NOT Exist in Code
These would cause serde deserialization failures or silent ignoring:
| Tool | Phantom Parameter | Notes |
|------|-----------------|-------|
| `m1nd_impact` | `depth` | Replaced by `direction` + `include_causal_chains` |
| `m1nd_warmup` | `task` | Correct name is `task_description` |
| `m1nd_scan` | `patterns` (array) | Correct is `pattern` (single string) |
| `m1nd_scan` | `top_k` | Correct is `limit` |
| `m1nd_timeline` | `node_id` | Correct is `node` |
| `m1nd_diverge` | `ref_a`, `ref_b` | Correct is `baseline` (single) |
| `m1nd_federate` | `label` (in repo object) | Correct is `name` |
| `m1nd_lock_create` | `center` | Replaced by `scope` + `root_nodes` |
| `m1nd_trace` | `stacktrace` | Correct is `error_text` |
| `m1nd_validate_plan` | `files` | Correct is `actions` (array of objects) |
| `m1nd_trail_save` | `notes` | Correct is `summary` |
| `m1nd_trail_merge` | `trail_id_a`, `trail_id_b` | Correct is `trail_ids` (array) |
| `m1nd_hypothesize` | `"uncertain"` verdict | Correct is `"inconclusive"` |
| `m1nd_hypothesize` | `"insufficient_data"` verdict | Does not exist |
| `m1nd_perspective_suggest` | `goal` | Does not exist |
| `m1nd_perspective_affinity` | `goal` | Does not exist |
| `m1nd_antibody_scan` | `"low"`, `"medium"`, `"high"` severity | Correct: `"info"`, `"warning"`, `"critical"` |
| `m1nd_antibody_create` | `id` (in pattern node) | Correct is `role` |
| `m1nd_antibody_create` | `label_pattern`, `match_mode` (in pattern node) | Correct is `label_contains` |
| `m1nd_antibody_create` | `from`/`to` strings in pattern edges | Correct is `source_idx`/`target_idx` integers |
| `m1nd_tremor` | `"Days7"`, `"Days30"` etc. | Correct: `"7d"`, `"30d"` etc. |
| `m1nd_trust` | `"TrustAsc"`, `"DefectsDesc"` etc. | Correct: `"trust_asc"`, `"defects_desc"` etc. |
| `m1nd_layers` | `"heuristic"` naming_strategy | Correct: `"auto"` |
| `m1nd_lock_watch` | `"OnAnyChange"` strategy | Correct: `"on_ingest"`, `"on_learn"`, `"manual"` |
## Output Schema Corrections
Beyond parameter inputs, the following output schemas were wrong:
| Tool | Field | Wrong | Correct |
|------|-------|-------|---------|
| `m1nd_ingest` | `files_processed` | wrong name | `files_scanned` + `files_parsed` |
| `m1nd_health` | `nodes` | wrong name | `node_count` |
| `m1nd_health` | `edges` | wrong name | `edge_count` |
| `m1nd_health` | `plasticity_edges` | does not exist | removed |
| `m1nd_activate` | `score` (in activated nodes) | wrong name | `activation` |
| `m1nd_activate` | `dimension_scores` | wrong name | `dimensions` |
| `m1nd_activate` | `ghost_edges[].from`, `.to`, `.confidence` | wrong names | `source`, `target`, `strength` |
| `m1nd_impact` | `blast_radius[{depth, nodes}]` | depth-bucketed | per-node with `node_id`, `signal_strength`, `hop_distance` |
| `m1nd_impact` | `total_affected`, `pct_of_graph`, `risk` | do not exist | replaced by `total_energy`, `max_hops_reached` |
| `m1nd_hypothesize` | `evidence[{path, hops}]` | wrong | `supporting_evidence`/`contradicting_evidence` with `type`, `description`, `likelihood_factor`, `nodes` |
| `m1nd_differential` | `structural_delta` array | does not exist | `new_edges`, `removed_edges`, `weight_changes`, `new_nodes`, `removed_nodes`, `coupling_deltas` |
| `m1nd_lock_diff` | flat fields | wrong | nested under `diff` object |
## Files Modified
- `/Users/cosmophonix/clawd/roomanizer-os/mcp/m1nd/.github/wiki/API-Reference.md` — comprehensive rewrite with all corrections applied
- `/Users/cosmophonix/clawd/roomanizer-os/mcp/m1nd/.github/wiki/VERIFICATION_REPORT.md` — this file (updated 2026-03-16: tool count 52→61, new tools section, verify=true feature documentation)
