Results for “evidence-ledger”

Use when sources must stay traceable across sessions and years, when each reference needs a passport recording where it was found, how it was verified, and where it is cited, or when a draft needs a cited-but-unverified sweep before submission.

Use when a study needs a preregistration or a frozen analysis plan, when confirmatory and exploratory analyses must be separated, or when every deviation from the registered plan must be logged before the result is reported.

Multi-dimensional lead qualification scoring. Evaluates leads against BANT criteria, firmographic fit, behavioral signals, and intent indicators. Outputs qualified/disqualified verdict with detailed reasoning.

Generate falsifiable trade strategy hypotheses from market data, trade logs, and journal snippets, with ranked hypothesis cards, experiment designs, kill criteria, and optional strategy.yaml export.

Designs incremental migration strategies, identifies service boundaries, produces dependency maps and migration roadmaps, and generates API facade designs for aging codebases.

Mines a codebase for evidence-backed improvement opportunities, forcing two escalation gates to surface breakthrough ideas, and outputs a ranked portfolio with a plan seed without implementing.

Create study guides with key concepts and review questions. TRIGGERS - Use when user needs help with study-guide related tasks.

Evidence-driven tracing lane that orchestrates competing tracer hypotheses in Claude built-in team mode

Clusters raw detection tickets into reusable edge concepts with thesis, invalidation signals, and strategy playbooks before strategy design.

Fact-forcing gate that blocks Edit/Write/Bash (including MultiEdit) and demands concrete investigation (importers, data schemas, user instruction) before allowing the action. Measurably improves output quality by +2.25 points vs ungated agents.

Derive selection thresholds from market data instead of hardcoding. Trigger when: (1) reviewing hardcoded parameters, (2) volume/price thresholds seem arbitrary, (3) selection returns too many/few candidates.

Runs an evidence-first, read-only workflow over large codebases via an external JSONL REPL server, retrieving only minimal file slices and greps to produce cited architecture and audit reports.