rana

rana/skills

Skill Claude CodeCodex

Inhabit a v3 agent cognitive architecture. Loads invocation, establishes identity, reads unread notes and SharedState. Use when entering a specific agent role.

rana/skills

Skill Claude CodeCodex

API surface consistency and design review. Use when reviewing REST endpoints, data models, response shapes, naming conventions, or pagination patterns.

rana/skills

Skill Claude CodeCodex

Cognitive archaeology — layered questioning that builds from surface inquiry through assumptions, perspectives, tensions, and meta-cognition to surface what conventional analysis misses. The sequence builds cognitive momentum; each layer prepares the ground for the next. Use for deep exploration of any subject…

rana/skills

Skill Claude CodeCodex

Question-driven autonomous research. Reads knowledge state, generates research questions, designs sub-agent prompts as questions. The meta-prompting protocol.

rana/skills

Skill Claude CodeCodex

Machine-readable skill inventory with composability properties, finding shapes, and stage mappings. For agent triage, programmatic selection, and toolkit self-awareness.

rana/skills

Skill Claude CodeCodex

Reasoning trace and correction. When thinking went wrong — yours or Claude's — trace back to find where pattern-matching replaced mechanism-tracing, where fear replaced exploration, or where compression happened prematurely. The cognitive equivalent of setting breakpoints in a thought process.

rana/skills

Skill Claude CodeCodex

Forward-propagation analysis. Given a proposed change, map 2nd and 3rd order effects across behavior, maintenance, user experience, and future optionality. Use before committing to significant decisions.

rana/skills

Skill Claude CodeCodex

Magic number and threshold audit. Extracts all numeric constants, checks cross-document consistency, and traces to source findings. Catches threshold mismatches, undocumented magic numbers, and constants without provenance.

rana/skills

Skill Claude CodeCodex

Multi-lens construction review. Runs one pass of the review protocol — spawns sub-agents for the appropriate wave, collects findings, reports convergence. Use after construction to audit fidelity, safety, and design quality. For the full protocol, see firm-scaffold/patterns/REVIEW.md.

rana/skills

Skill Claude CodeCodex

Rapid reorientation to a different area of the codebase. Delivers a 60-second briefing on a subsystem's current state, recent changes, key files, and known issues. Use when switching between tasks.

rana/skills

Skill Claude CodeCodex

Iteration gate. Measures delta between passes, declares stability or directs the next cycle.

rana/skills

Skill Claude CodeCodex

Cross-repository consistency audit for related repositories (e.g., spec + scaffold + archive). Detects mismatches in cross-references, file paths, counts, thresholds, and shared terminology across repo boundaries.

rana/skills

Skill Claude CodeCodex

Multi-perspective validation gate. Stress-test claims through independent reasoning paths before promotion.

rana/skills

Skill Claude CodeCodex

Stress-test through attachment — build genuine appreciation, then break what you built, then rebuild from what survives. The only skill whose trajectory passes through zero.

rana/skills

Skill Claude CodeCodex

Simplification and editorial sharpening. Use when designs feel heavy, over-engineered, or accumulated rather than composed. The editorial knife.

rana/skills

Skill Claude CodeCodex

Inhabit a specific cultural, spiritual, or demographic perspective and audit the design for sensitivity, inclusion, accessibility, and blind spots. Use when checking how the portal serves specific populations.

rana/skills

Skill Claude CodeCodex

Broad hypothesis-free quality sweep. Combines coherence, gaps, and errors in a single pass. Use when you don't have a specific concern — for targeted analysis, compose specific skills instead (e.g., drift-detect for architecture, reframe + gaps for opportunities).

rana/skills

Skill Claude CodeCodex

Unified document health check combining identifier audit, omission search, and cross-document consistency in a single pass. Reads all documents once, threads findings across phases. Use for recurring document maintenance or pre-implementation review.

rana/skills

Skill Claude CodeCodex

Documentation practice quality review. Use when evaluating whether docs serve AI readers (Claude), human developers, and stakeholders effectively.

rana/skills

Skill Claude CodeCodex

Detect unconscious architectural evolution. Compare the stated architecture in documents against emergent patterns in the actual code. Surface drift that nobody designed. Use periodically or when the codebase feels inconsistent.

rana/skills

Skill Claude CodeCodex

FTR corpus curation — compare, merge, split, archive, and triage specification files. The operational layer between spec-survey diagnosis and manual refactoring. Write-enabled (preview by default).

rana/skills

Skill Claude CodeCodex

Corpus freshness audit. Systematic staleness detection across state accuracy, content drift, and relevance decay. The triage pass that identifies which FTRs need verify, update, or archive.

rana/skills

Skill Claude CodeCodex

Systematic search for omissions, blind spots, and missing decisions. Use when preparing for implementation, after completing a design phase, or when asking "what am I not seeing?".

rana/skills

Skill Claude CodeCodex

Diagnose the current state of strategy space geometry — dimensions, holes, grades, eigenvalues, clusters. Use weekly or when the manifold feels stale.