Project Spike

Project Spike

Use this skill only when explicitly invoked as $project-spike. It creates or updates spike artifacts at docs/spikes/<topic>.md and tracks open/resolved spike entries in progress.txt.

Core Rules

• Read progress.txt fresh from disk before acting.
• Require a bootstrapped project. If progress.txt is missing, tell the user to run $project first.
• Run research before red-team review in separate Codex sub-agent contexts. Do not silently collapse the research and red-team passes into one local context. If sub-agent delegation is unavailable, stop and tell the user the spike cannot preserve the required adversarial isolation in this environment.
• Preserve original spike findings during follow-up; append dated follow-up entries.
• Ask concise chat questions for user choices; use structured input tools when available.
• Do not invoke the next skill automatically. Recommend the next explicit skill.
• Read all referenced spike workflow files before creating or updating an artifact; they contain required audit-trail and follow-up rules.

Workflow

1. Gather the research question and available tooling from the user request or ask concise follow-up questions.
2. Read references/spike-format.md and generate the topic slug.
3. If docs/spikes/<slug>.md exists, enter follow-up mode; otherwise create a new spike.
4. Read references/research-agent.md for the research pass and references/redteam-agent.md for the red-team pass.
5. Assemble or update the artifact using references/spike-format.md.
6. Update progress.txt using references/progress-format.md. New spikes are [ ]; resolved spikes are [x] ... Resolved: <YYYY-MM-DD>.

Reference Files

• references/spike-format.md
• references/research-agent.md
• references/redteam-agent.md
• references/progress-format.md

Completion

Report the spike artifact path, state update, follow-up count if any, and recommend $project for status.