gsd-skill-creator

Extract creative intent from images into executable build specs. Activates on images + build intent, "image to mission", "i2m", or capturing visual energy in code/design.

Unit Circle Laboratory + sc-dev-team autonomous mission control. Provides human-in-the-loop automation layer, pipeline management, context lifecycle, stuck-state prevention. UC re-execution series and dev branch milestones. Adapted from uc-lab pattern for code plan/execute/verify/complete. Auto-activates during UC milestone work; trigger: user says 'bring up the dev team'.

Generate intelligence briefings for the planning dashboard. Use this skill whenever a request file appears in `.planning/console/inbox/pending/` whose `type` field starts with `intelligence.` (refresh_briefing, triage_finding, snapshot_diff, investigate_section, dismiss_finding). The skill reads the per-project KB at `.gsd/intelligence/intelligence.db`, synthesizes a briefing with a causal hypothesis + acknowledged uncertainty + confidence label and ranked moves, then writes the result back to the KB. Always trigger this skill for these request types — do not generate briefings manually.

Sketched Isotropic Gaussian Regularization primitive. Scalar loss matching the embedding distribution to a standard-normal target via Cramér-Wold slicing and the Epps-Pulley empirical characteristic function test. Port of rbalestr-lab/lejepa (MIT). Default-off in v1.49.571.

Adversarial spec-compliance PR review — cross-references diffs against approved specs, verifies runtime claims against source, detects competing PRs, audits scope/convention compliance. Use before merging.

Pick among candidate outputs (code, configs, plans) by running them on diverse inputs and clustering by behavioural fingerprint, rather than by textual aggregation or log-probability. Activates when an executor returns multiple plausible candidates that need disambiguation, when output-majority voting would be the default choice, or when reviewing generated code that has not yet been validated. The 2026 evidence (Semantic Voting, arxiv 2605.08680v1) is that any execution-based selector dominates output-majority voting by 19-52pp; sketch-generated inputs beat random fuzz by 11.3pp. Triggers: "pick the best candidate", "majority vote on code", "select from N samples", "validate the generated output", "behavioural verification".

Classify the information-need of a query and dispatch it to the appropriate retrieval or reasoning strategy. Use before read-side memory access, before multi-strategy retrieval, or any time you'd otherwise default to "one retriever for everything". Returns a strategy label, a token budget, and a retrieval depth so downstream handlers can be specialised. Backed by Pre-Route (arxiv 2605.10235v2) and MemFlow (arxiv 2605.03312v1), which together show LLMs possess latent routing ability elicitable via a structured prompt — and that externalising the routing decision improves small-model performance by ~2x. Triggers: "route this", "what strategy", "before retrieving", "intent classification", or any query whose ideal handling depends on what KIND of question it is.

Audit a skill by running a paired probe — the same task once with the skill loaded and once without — segment both traces into goal-directed phases, align phases, and emit a SIP report (surface anchoring, template copy, excess planning, task recovery, off-task artifact). Use whenever a skill is created, modified, or proposed for retirement. Pass-rate is BLIND to most skill effects: CTA (arxiv 2605.11946v1) shows a single skill can produce 522 measurable behavioural changes across 49 tasks while pass-rate moves only +0.3%. Triggers: "audit this skill", "is this skill helping", "retire skill", "before shipping skill", "behavioural impact of skill X", or any skill review event.

Manages skill-creator integration with GSD workflows including skill loading, session observation, bounded learning guardrails, and pattern detection. Use this skill whenever: executing any GSD phase (to ensure relevant skills are loaded first), starting or resuming a work session (to check for pending suggestions), the user asks about skills, patterns, or skill-creator status, performing code changes that might represent repeating patterns, the user corrects Claude's output (highest-signal observation), or when skill refinement, creation, or suggestion review is discussed. Critical for maintaining the adaptive learning layer.

Before dispatching a multi-agent team, run a spectral diagnostic on the proposed communication graph and emit a (ρ, Δ, κ) coordination signature plus a pass/fail verdict against per-task-class thresholds. Builds the row-stochastic operator P from the team graph, computes the successor representation M = (I − γP)⁻¹, and ranks the topology for robustness (κ, condition number), consensus (Δ, spectral gap), and drift (ρ, spectral radius). Per Parks & Alharthi (arxiv 2605.11453), rank order on (κ, Δ, ρ) predicts coordination quality pre-execution with rank correlations of 1.0 / 0.5 / −1.0. Triggers: "dispatch the team", "team topology check", "before running the agents", "is this team configuration OK", "team pre-flight".

Token budget tracking and enforcement for Gastown convoy-level execution. Hard limits with pre-execution checking, per-convoy and per-agent tracking, structured stop reasons.

Interrupt controller converting polled to proactive agent execution. Per-runtime strategies, configurable thresholds, Deacon heartbeat supervision. Fights LLM passivity bias.

Cascade the same edit pattern across N files safely. Use when applying the same refactor to multiple files (e.g. swap import paths across 11 scripts, rename a symbol, migrate a call signature). Detects the common-shape-across-files situation and turns an N-file cascade into a planned audit → apply → verify workflow instead of N sequential manual edits.

Persist progress for long-running jobs (batched LLM calls, large ingestions, multi-hour syncs) so that a context reset, crash, or interrupt doesn't lose work. Use whenever a job iterates over N items and completing item K matters independently. Provides a resumable.mjs library pattern plus the skill's invocation heuristics.

Triggers for automatically invoking the decision-framework skill before irreversible or high-blast-radius operations. Use as a checkpoint: when a proposed action matches a trigger pattern, pause and run the appropriate framework (first-principles, inversion, decision matrix, second-order thinking) before proceeding. Prevents ad-hoc calls from becoming regret.

Emit Postgres.sql and SQLite.sql from a single schema spec so tools work across both drivers without duplicating DDL by hand. Use when designing a schema that needs to support both shared Postgres deployments and zero-config SQLite. Reduces two-file sync burden to a single source edit.

Conversational guide for using tools/session-retro/observe.mjs to capture friction, wins, corrections, decisions, gaps, tool-use, and checkpoints as they happen during a session. Use at session start (to kick off the log), at inflection points (to record events), and at session end (to archive and feed the retrospective generator).

Use when the user asks to forge, scaffold, or author a new cartridge — content cartridge, department cartridge, or chipset bundle. Drives the cartridge-forge loop (scaffold → fill → gate checks → commit) via the skill-creator cartridge CLI.

Calculus, series, numerical methods, and optimization for computational mathematics. Covers limits (epsilon-delta, L'Hopital, squeeze theorem), derivatives (rules, optimization, related rates), integrals (Riemann sums, Fundamental Theorem of Calculus, integration techniques), series (convergence tests, Taylor and Maclaurin expansions), differential equations (separable, linear, exact), numerical methods (Newton-Raphson, Euler method, Simpson's rule), and computational pitfalls (floating-point arithmetic, catastrophic cancellation, numerical stability). Use when computing derivatives, integrals, limits, series, solving differential equations, or analyzing numerical accuracy.

Security hygiene for GSD's self-modifying skill and agent system. Use this skill whenever: creating, editing, or deleting skill files (.claude/skills/, .claude/commands/), modifying agent definitions (.claude/agents/), working with YAML configuration or chipset files, handling JSONL observation data (.planning/patterns/), processing community-contributed skills or chipsets, any file path operations that could involve user input, or when installing/updating project-claude configuration. Also activates for discussions about skill-creator security, trust models, or content hygiene.

Token budget tracking and enforcement for Gastown convoy-level execution. Hard limits with pre-execution checking, per-convoy and per-agent tracking, structured stop reasons.

Pipeline retirement protocol for completed work items. Implements a 7-stage validate-commit-push-submit-notify-cleanup-terminate pipeline with irreversibility after push. Done means gone.

Instruction dispatch pipeline for routing work items to available agents. Implements a 7-stage fetch-allocate-prepare-hook-store-launch-confirm pipeline with batch convoy mode, formula expansion, idempotent dispatch, and crash recovery.

REST API design best practices. Use when designing APIs, choosing status codes, or creating endpoints.

Cascade the same edit pattern across N files safely. Use when applying the same refactor to multiple files (e.g. swap import paths across 11 scripts, rename a symbol, migrate a call signature). Detects the common-shape-across-files situation and turns an N-file cascade into a planned audit → apply → verify workflow instead of N sequential manual edits.

Use when the user asks to forge, scaffold, or author a new cartridge — content cartridge, department cartridge, or chipset bundle. Drives the cartridge-forge loop (scaffold → fill → gate checks → commit) via the skill-creator cartridge CLI.

Persist progress for long-running jobs (batched LLM calls, large ingestions, multi-hour syncs) so that a context reset, crash, or interrupt doesn't lose work. Use whenever a job iterates over N items and completing item K matters independently. Provides a resumable.mjs library pattern plus the skill's invocation heuristics.

Reviews code for bugs, style, and best practices. Use when reviewing PRs or checking code quality.

Crafts and generates professional conventional commit messages following Angular/Conventional Commits format. Use when committing changes, writing git commit messages, or when user mentions commit message.

Creates context handoff documents for session continuity. Use when ending sessions, switching tasks, or handing off work.

Triggers for automatically invoking the decision-framework skill before irreversible or high-blast-radius operations. Use as a checkpoint: when a proposed action matches a trigger pattern, pause and run the appropriate framework (first-principles, inversion, decision matrix, second-order thinking) before proceeding. Prevents ad-hoc calls from becoming regret.

Thinking frameworks for decisions and problem analysis. Use when evaluating options, root causes, or prioritizing.

Environment configuration and secrets management. Use when setting up .env files, managing secrets, or configuring environments.

Safe file operation patterns. Use when performing bulk file operations or writing deployment scripts.

GSD tutorial, command reference, and workflow explainer. Explains what a GSD workflow command does before you run it. Activates when user is new to GSD, asks about commands, wants to understand workflows, or mentions preview.

Use when GSD commands are failing or .planning/ artifacts look inconsistent. Run before any GSD execute-phase or verify-work attempt to catch cross-artifact mismatches early.

Traces decisions and requirements through GSD artifacts. Use when user asks "why did we...", "what happened to...", or mentions decision history.

Use when the user asks "what should I work on", mentions GSD phases, ROADMAP, or .planning/ artifacts. Routes to the correct GSD command for the current project state.

Emit Postgres.sql and SQLite.sql from a single schema spec so tools work across both drivers without duplicating DDL by hand. Use when designing a schema that needs to support both shared Postgres deployments and zero-config SQLite. Reduces two-file sync burden to a single source edit.

Package conversation research into a GSD-ready mission package. Produces a three-stage Vision → Research → Mission pipeline as a LaTeX PDF following GSD/NASA SE methodology, designed to be handed to gsd-skill-creator for execution. Use this skill when the user has been discussing, researching, or brainstorming a topic and then asks to turn it into a research mission, research pack, mission package, or says 'package this as a mission', 'make this a research pack', 'turn this into a mission for skill-creator', 'create a research mission from this', or 'use the research mission skill'. The skill harvests findings, sources, and structure from the current conversation and any prior research, then produces the complete pipeline document. Also trigger if the user asks to 'create a research mission on [topic]' cold — in that case, conduct web research first, then package.