Skip to content

Investigations

Time-boxed research spikes and validation experiments. Use an investigation doc to answer a specific question before committing to a design or implementation — e.g. proving a library can handle a requirement, reproducing a bug, or validating a performance assumption.

Design docs, plans, and implementation docs can reference investigations by ID (e.g. INV-0001) to document how open questions were resolved.

IDTitleStatusDateAuthorLink
INV-0001OpenAI and OpenCode skill/plugin format compatibility with ClaudeOpen2026-04-18Donald Gifford0001-openai-and-opencode-skill-and-plugin-format-compatibility-with.md
INV-0002Ralph-loop permission matcher bug and fork planConcluded2026-04-20Donald Gifford0002-ralph-loop-permission-matcher-bug-and-fork-plan.md
INV-0003Phase 1.8 dogfood findings on external Claude pluginsClosed2026-04-21Donald Gifford0003-phase-18-dogfood-findings-on-external-claude-plugins.md
INV-0004Donald-loop stop hook silently fails promise detection on transcripts with unescaped newlinesClosed2026-04-21Donald Gifford0004-donald-loop-stop-hook-silently-fails-promise-detection-on.md
INV-0005Phase 2 dogfood findings: marketplaces, MCP, and spec divergenceClosed2026-04-23Donald Gifford0005-phase-2-dogfood-findings-marketplaces-mcp-and-spec-divergence.md
INV-0006Rule coverage audit against current Claude Code docsConcluded2026-07-09Donald Gifford0006-rule-coverage-audit-against-current-claude-code-docs.md