dotfiles

Track agent decisions, metrics, token costs, and generate reports. Use when logging task execution, tracking agent performance, or generating observability reports.

Unified ai-brain skill combining memory operations and KB query operations with promotion flow.

Bootstraps a Jira hierarchy snapshot before any multi-issue create / edit session. Disambiguates the parent input (issue key, spec URL, Confluence page URL, or `greenfield`), walks parents up (depth-cap 4), walks children down via JQL recursion (depth-cap 4, max 50 children per node), and returns a structured tree alongside the issue-type scheme and link-type metadata the caller needs to render a draft. Used primarily by `atlassian-pm`; secondary consumers (after the G2.5 retrofit) are the upstream Atlassian plugin skills `spec-to-backlog` and `capture-tasks-from-meeting-notes`. The skill **does not render**; rendering belongs to the caller (typically `atlassian-pm` rendering a markdown tree to chat).

Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full, ultra. Defaults: main chat = lite, subagents = ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.

Generate → execute → fail → analyze → fix → repeat cycle with failure pattern recognition. Use when implementing changes that need verification and automatic recovery.

Heuristic context budget checks using file-size and count proxies — integrates with context-cache-discipline; no tokenizer introspection claims.

Token and prompt-cache discipline — lean always-on rules, stable prefixes, skill indirection, minimal Task payloads, bounded reads.

Two-pass plan-critic loop (`cro`); ledger schema, adversarial rubric, bounce rubric, caps, observability stages.

Agent-based project identity derivation for Knowledge Base. Reads git config directly without shell commands. Worktree-safe — all worktrees of the same repo derive the same identity.

Standardizes how `n8n-builder` enters and runs Stage C (implementation) with worker fan-out, group checkpoints, and gated progression. Use when n8n-builder transitions from CRO `approved`/`degraded_skip` to execution.

Standardizes how `n8n-builder` performs planning triage, specialist consultation, and produces plan v0 plus post–v0 edit round; mandatory `cro-loop` runs at execution intent before Stage C. Use when n8n-builder is in Stage A (planning).

Three-level parallelism (phase group → phase-task → intra-role fan-out), parallelize-by-default stance, disjoint-touches pre-flight, merge/rollback for horizontal fan-out, failure semantics, and dynamic concurrency caps for orchestrators consuming a cto plan.

Declarative pipeline execution with stages, conditions, data passing, and rollback. Use when executing multi-stage workflows.

Validates generated output before writing to disk. Use when implementing changes that need pre-write verification to catch errors early.

Standardizes how `remotion-builder` runs Stage C (implementation) after tech CRO approval — remotion CLI, browser ensure, Skia smoke, ffmpeg from approved recipes, group checkpoints.

Standardizes how `remotion-builder` performs planning triage, specialist consultation, and produces plan v0; mandatory tech `cro-loop` at execution intent before Stage C. Use when remotion-builder is in Stage A (planning).

Programmatic rule enforcement with glob patterns, priority resolution, and pre/post validation. Use when checking compliance before or after agent actions.

Protocol for subagent → parent responses: single fenced YAML envelope, strict caveman-ultra for compressed fields, verbatim for paths/errors/code. Applies to every `Task` invocation. Use when spawning subagents or parsing their output.

Automated task classification, agent selection, and state tracking. Use when routing tasks to agents, selecting pipelines, or managing task state.

Discovery + naming convention reference for typed dev/SME/QA/devops team members in any workspace folder. Primary consumer: `tech-lead` (org-tier).

Two-pass plan-critic loop (`cro`); ledger schema, adversarial rubric, bounce rubric, caps, observability stages.

Heuristic context budget checks using file-size and count proxies — integrates with context-cache-discipline; no tokenizer introspection claims.

Token and prompt-cache discipline — lean always-on rules, stable prefixes, skill indirection, minimal Task payloads, bounded reads.

Protocol for subagent → parent responses: single fenced YAML envelope, strict caveman-ultra for compressed fields, verbatim for paths/errors/code. Applies to every `Task` invocation. Use when spawning subagents or parsing their output.

Three-level parallelism (phase group → phase-task → intra-role fan-out), parallelize-by-default stance, disjoint-touches pre-flight, merge/rollback for horizontal fan-out, failure semantics, and dynamic concurrency caps for orchestrators consuming a cto plan.

Use when deciding whether to extract a shared utility, create a base class, build a framework layer, or generalize a pattern -- especially when the pattern has fewer than three concrete instances or the abstraction adds cognitive overhead without clear payoff.

Use when delegating tasks to AI agents, designing prompts for structured output, validating AI-generated code or reasoning, detecting hallucination, or when AI output needs constraint checking and human oversight before integration.

Bootstraps a Jira hierarchy snapshot before any multi-issue create / edit session. Disambiguates the parent input (issue key, spec URL, Confluence page URL, or `greenfield`), walks parents up (depth-cap 4), walks children down via JQL recursion (depth-cap 4, max 50 children per node), and returns a structured tree alongside the issue-type scheme and link-type metadata the caller needs to render a draft. Used primarily by `atlassian-pm`; secondary consumers (after the G2.5 retrofit) are the upstream Atlassian plugin skills `spec-to-backlog` and `capture-tasks-from-meeting-notes`. The skill **does not render**; rendering belongs to the caller (typically `atlassian-pm` rendering a markdown tree to chat).

Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full, ultra. Defaults: main chat = lite, subagents = ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.

Use when writing architecture decision records, documenting trade-offs, drafting design proposals, writing postmortems, summarizing technical reasoning, or when a decision lacks documented context and clear justification.

Generate → execute → fail → analyze → fix → repeat cycle with failure pattern recognition. Use when implementing changes that need verification and automatic recovery.

Structured feedback propagation between pipeline stages. Use when review stages produce findings that should trigger re-implementation.

Dev–Reviewer–QA closed-loop orchestration for implementation tasks — sequence, escalation, retries, merged-diff fan-in, loop state tracking. Invoke when coordinating tech-lead or code-reviewer workflows that chain dev-, reviewer-, and qa-* agents or senior-dev fallback.

Use when estimating infrastructure costs, comparing architectural approaches by cost, evaluating managed service trade-offs, reviewing cloud pricing models, or when a design decision has unclear cost implications or vendor lock-in risk.

Use when designing for graceful degradation, implementing retry or circuit breaker logic, handling partial failures, evaluating idempotency, setting up dead-letter queues, or when a system lacks resilience under dependency failure or network instability.

Agent-based project identity derivation for Knowledge Base. Reads git config directly without shell commands. Worktree-safe — all worktrees of the same repo derive the same identity.

Use after completing major features, recovering from incidents, making architectural decisions that turned out wrong, or when repeated mistakes suggest stagnant mental models and missing feedback loops.

Standardizes how `n8n-builder` enters and runs Stage C (implementation) with worker fan-out, group checkpoints, and gated progression. Use when n8n-builder transitions from CRO `approved`/`degraded_skip` to execution.

Standardizes how `n8n-builder` performs planning triage, specialist consultation, and produces plan v0 plus post–v0 edit round; mandatory `cro-loop` runs at execution intent before Stage C. Use when n8n-builder is in Stage A (planning).

Use when defining metrics before implementation, designing structured logging, setting SLO targets, instrumenting tracing boundaries, building health checks, or when an outage reveals missing signals and blind spots in operational visibility.