spike-workflow
Apply when the right APPROACH is unknown and must be discovered before committing — a hard algorithm or big-O/rendering decision, a complex state machine, tricky native/platform integration, a gnarly gesture/animation problem, a performance optimization, anything where you would otherwise guess an approach and build it. Use this whenever the task is "figure out how to...", "what is the best way to...", "optimize X", "this is slow/complex", "spike on Y". The dominant failure mode is COMMITTING TO THE WRONG APPROACH and DECLARING FALSE VICTORY against a weak baseline (agent psychosis). The rule is: establish the FLOOR (theoretical limit / reference) FIRST and judge every result against it, enumerate solution CLASSES (consult the mental-models skill) rather than optimizing within one, benchmark, then collapse the spike into a feature or refactor. Delegates to feature-workflow's change-kind gate-runners with --kind spike. Do NOT use once the approach is known — collapse and run that kind's loop.
What this skill does
# Spike Workflow
Resolve an unknown *approach* before committing to one. The unknown is which
*class* of solution is right — and the trap is that you cannot search for a class
you do not know exists ("you don't look for what you don't know"). So the parse
is a **benchmark against a FLOOR**, plus a **solution-class enumeration** driven
by transferable reframings.
## Why the FLOOR is non-negotiable (the anti-psychosis gate)
A result is only good against the *limit*, never against your own starting point.
The cautionary case: an agent drove a renderer 88ms → 1.5ms (150K → ~500 allocs)
and declared triumph; a hand-written reference did the same work in ~20µs with 0
allocations — a 75× gap it could not see because it compared to its own naive
baseline, not the floor. **Trusting "good" without parsing it against the limit
is the quality twin of trusting unparsed input.** Estimate the floor without
knowing the winning approach: big-O lower bound, zero-allocation ideal, memory
bandwidth ceiling, a hand-rolled reference. Then a result that sits far from the
floor *forbids* declaring victory — the unknown-unknown becomes a visible anomaly.
## The loop (the feature-workflow spine, spike instance)
1. **STEP 0 classified this as a spike.** `change_new.py --kind spike --name "..."`.
2. **Establish the FLOOR first** and write it into the manifest. (See
`references/floor.md`.)
3. **Enumerate solution CLASSES, not variants.** Consult the `mental-models`
skill: feed it the smell ("this is slow", "this is O(n^2)", "deep nesting")
and it returns the reframing questions that *open classes you would not have
searched for* (allocation-bound vs compute-bound? data-oriented layout?
asymptotically better structure? mechanical sympathy?).
4. **Gate:** `change_check.py --kind spike <manifest>` — blocks unless a floor is
declared AND results are expressed as a ratio to it.
5. **Benchmark across classes**, report best-vs-floor as a ratio.
6. **Collapse:** the chosen class turns the spike into a feature (new behavior),
a refactor (behavior-preserving restructure), or a perf-refactor (output
identical, timing changed → needs BOTH a characterization net and a
benchmark-vs-floor gate). Run that kind's loop next.
7. **Feed the learning loop:** any gap that exposed a *missing* model — a class
you didn't think to consider — is recorded in `mental-models` with its trigger
smell. The gap is the authority. That is how unknown-unknowns become known.
## The one law here
A spike produces a *decision backed by a floor-anchored benchmark*, not code you
keep. Resist building the first plausible approach; the cost of the wrong class
dwarfs the cost of the spike.
## Files
- `references/floor.md` — estimating a floor without knowing the answer; the
reframing-checklist handshake with mental-models.
Related in AI Agents
skill-development
IncludedComprehensive meta-skill for creating, managing, validating, auditing, and distributing Claude Code skills and slash commands (unified in v2.1.3+). Provides skill templates, creation workflows, validation patterns, audit checklists, naming conventions, YAML frontmatter guidance, progressive disclosure examples, and best practices lookup. Use when creating new skills, validating existing skills, auditing skill quality, understanding skill architecture, needing skill templates, learning about YAML frontmatter requirements, progressive disclosure patterns, tool restrictions (allowed-tools), skill composition, skill naming conventions, troubleshooting skill activation issues, creating custom slash commands, configuring command frontmatter, using command arguments ($ARGUMENTS, $1, $2), bash execution in commands, file references in commands, command namespacing, plugin commands, MCP slash commands, Skill tool configuration, or deciding between skills vs slash commands. Delegates to docs-management skill for official documentation.
reprompter
IncludedTransform messy prompts into well-structured, effective prompts — single or multi-agent. Use when: "reprompt", "reprompt this", "clean up this prompt", "structure my prompt", rough text needing XML tags and best practices, "reprompter teams", "repromptception", "run with quality", "smart run", "smart agents", multi-agent tasks, audits, parallel work, anything going to agent teams. Don't use when: simple Q&A, pure chat, immediate execution-only tasks. See "Don't Use When" section for details. Outputs: Structured XML/Markdown prompt, quality score (before/after), optional team brief + per-agent sub-prompts, agent team output files. Success criteria: Single mode quality score ≥ 7/10; Repromptception per-agent prompt quality score 8+/10; all required sections present, actionable and specific.
adaptive-compaction
IncludedAdaptive add-on policy and recovery layer that decides WHEN to compact, prune, snapshot, or fork -- replacing fixed-percent auto-compaction across Claude Code, Codex, and MCP-capable hosts. Trigger on auto-compact timing or damage: "when should I compact", "is it safe to compact now or start a fresh session", "auto-compact fires too early/mid-task", "switching to an unrelated task but the window still has space", "context rot", "answers get worse the longer the session runs", "the agent forgot the plan or my decisions after it summarized", "add a layer on top that manages context without changing the agent", raising autoCompactWindow to give the policy room, or installing/tuning a cross-tool compaction policy or PreCompact hook -- even when "compaction" is never said but the problem is context-window pressure or post-summarization memory loss. Do NOT use to summarize a conversation, build RAG, write a summarization prompt (decides WHEN not HOW), or answer max-context-length trivia.
agent-skill-creator
IncludedCreate cross-platform agent skills from workflow descriptions. Activates when users ask to create an agent, automate a repetitive workflow, create a custom skill, or need advanced agent creation. Triggers on phrases like create agent for, automate workflow, create skill for, every day I have to, daily I need to, turn process into agent, need to automate, create a cross-platform skill, validate this skill, export this skill, migrate this skill. Supports single skills, multi-agent suites, transcript processing, template-based creation, interactive configuration, cross-platform export, and spec validation.
llm-wiki
IncludedUse when building or maintaining a persistent personal knowledge base (second brain) in Obsidian where an LLM incrementally ingests sources, updates entity/concept pages, maintains cross-references, and keeps a synthesis current. Triggers include "second brain", "Obsidian wiki", "personal knowledge management", "ingest this paper/article/book", "build a research wiki", "compound knowledge", "Memex", or whenever the user wants knowledge to accumulate across sessions instead of being re-derived by RAG on every query.
skill-master
IncludedAgent Skills authoring, evaluation, and optimization. Create, edit, validate, benchmark, and improve skills following the agentskills.io specification. Use when designing SKILL.md files, structuring skill folders (references, scripts, assets), ingesting external documentation into skills, running trigger evals, benchmarking skill quality, optimizing descriptions, or performing blind A/B comparisons. Keywords: agentskills.io, SKILL.md, skill authoring, eval, benchmark, trigger optimization.