TDD log files and sidecar-to-log cross-validation.

TDD sidecar JSON (benchmarks 14, 41).

Terminal Gate section in PROGRESS.md.

Test file extension matches project language (benchmark 47).

Use case identifier section (benchmarks 43, 48).

§10 invariants #7, #12 — disposition completeness + legal fix_type × disposition.

§10 invariant #9 — reference_docs/cite/ contains only .txt/.md. v1.5.2 collapsed the old formal_docs/+informal_docs/ split into a single refe

Dispatcher that runs every Layer-1 mechanical check from schemas.md §10.

§10 invariant #10 — quality/INDEX.md exists with all §11 required fields. v1.5.4 Part 1 / Round 1 Council finding C2-1 + Round 2 Council find

§10 invariant #13 — manifest wrapper shape. Four record-shaped manifests (formal_docs / requirements / use_cases / bugs) use `records`; citat

§10 invariants #1, #4, #8, #11, #14 — REQ shape, citation gating, functional_section.

§10 invariant #17 — Council-of-Three majority-overreaches rule. Layer-2 semantic check (Phase 6). Gate does NOT re-run the semantic review; i

v1.5.2 Lever 3: Phase 5 cardinality reconciliation gate. Surfaces every failure from validate_cardinality_gate() as a fail() entry.

Phase 3b BLOCK-4 cross-reference + DQ-5 structural validation. Validates quality/phase3/pass_d_council_inbox.json against the DQ-5 schema AND

schemas.md §10 invariant #22 — BUG.divergence_type on v1.5.3-shaped manifests. Legacy manifest: one WARN, then skip. v1.5.3-shaped: every BUG

schemas.md §10 invariant #23 — FORMAL_DOC.role on v1.5.3-shaped manifests. Legacy manifest (no v1.5.3 fields anywhere): one WARN, then skip.

schemas.md §10 invariant #21 (second part) — skill_section consistency. On a v1.5.3-shaped requirements manifest, REQs with source_type == 's

schemas.md §10 invariants #21 (first part) — REQ.source_type presence. Legacy manifest: one WARN, then skip. v1.5.3-shaped: every REQ MUST ha

Version stamp consistency (benchmark 26). Returns detected skill_version.

Bug writeups section (benchmark 30).

Clamp a requested step budget into a sane range.

Build a complete analysis document.

Convert a document to PDF.

Extract text from a source document (for the AI to read before writing analysis).

Generate a single screenshot from a PDF.

Check if all dependencies are available.

Collect code metrics: file counts by extension, total LOC.

Compare two elements and return (simple_diffs, nested_set_attrs). simple_diffs: differences in non-Set attributes nested_set_attrs: list

Compare column names, types, nullability. Return drift report and common columns.

* Recursively copy a directory.

Count total images across all slides.

Count bugs in list that have `field` set.

Count source files up to 4 dirs deep, excluding vendor/node_modules/etc.

Create a full-screen dark backdrop whose alpha is controlled by the settings slider.

Create an arrow element. Args: from_x: Starting X coordinate from_y: Starting Y coordinate to_x: Ending X coordi

Create a CopilotClient that works in both normal and frozen builds.

Create a text label element. Args: text: Label text x: X coordinate y: Y coordinate Returns: Te

Show classification working on sample outputs.

Detect CI/CD pipeline configurations.

Detect containerization and orchestration configs.

Detect monorepo signals.

Detect performance testing and profiling markers.

Walk up to 3 dirs deep, return first language whose extension is present. Mirrors bash `find -maxdepth 3 -not -path ...` behavior.

Detect security and compliance configurations.

Determine exit code based on analysis results.

Filter abnormal months - overall statistics, not by organization

Drop high correlation features - based on original gain, drop one feature at a time Args: data: Data features: Feature list

Drop high missing rate features

Null Importance to remove high noise features

Drop high PSI features - by organization + month-by-month version Multi-processing at feature level, loop through organizations, parallel cal

Drop low IV features - multi-process version, returns IV details and IV processing table Args: overall_iv_threshold: Overall IV thres

* Ensure the output directory exists

Inject certifi's CA bundle into the environment. On macOS frozen builds the system certificate store is unreachable, so the CLI subprocess fa

Export cleaning report to xlsx - one sheet per step Args: filepath: Output path steps: Cleaning step list [(step name, DataFr

* Extract file extension from a glob pattern

* Extract full MCP server configs from an agent file * @param {string} filePath - Path to the agent file * @returns {Array<{name:string,type?:string

Extract raw image bytes and content type from a shape.