oleksify
User[Mirror] https://tangled.org/oleksify.me/rnd-framework
Categories
Indexed Skills (62)
rnd-design
Use when exploring architectural alternatives before planning — generates 2-3 design options with trade-off evaluation, produces a design-spec.md artifact, and gates on explicit user approval before planning proceeds
rnd-experiments
Use when independently verifying built work — defines how verifiers write experiment tests from the spec alone before reading Builder code, preventing false PASSes through mandatory independent validation
rnd-local-experts
Use when discovering and surfacing project-local agents and skills from the target project's .claude/ directory — scan, read frontmatter, and present a structured summary for the Planner to reference in pre-registrations
rnd-verification
Use when independently verifying built work against pre-registered criteria — information-barrier verification with evidence-based verdicts and failure mode analysis
bash-hook-testing
Use when writing or modifying tests for rnd-framework hook scripts — covers the test-helpers.sh framework, run_hook pattern, assertions, environment mocking, and test organization
bun-scripting
Use when writing helper scripts, utility tasks, or automation within the R&D pipeline — prefer Bun (TypeScript) over Python when available
code-review
Use when reviewing code changes for quality, correctness, and security — defines review categories, severity levels, verdicts, and the structured report format
committing
Use when creating git commits — enforces commit message style, length limits, and user confirmation before committing
fp-practices
Use alongside KISS practices to guide agents toward functional programming patterns — pure functions, data transformations, composition, command-query separation, and immutability
hook-authoring
Use when writing or modifying rnd-framework hook scripts — covers hook anatomy, exit code protocol, stdin parsing, fast-path patterns, and hooks.json registration
lib-sh-patterns
Use when writing hooks or lib scripts — covers lib.sh shared utilities, path predicates, response functions, stdin parsing, and the active_session_dir caching pattern
outside-view
Use in Phase 1 of rnd-start (after the premortem fan-out, before the Planner spawn) to inject a historical reference-class block — per-shape FAIL rates from the session corpus — into the Planner's context as a calibration anchor.
plugin-architecture
Use when working with the plugin structure — covers config dir detection, path matching, hooks.json, and hook event availability
plugin-versioning
Use when bumping plugin version, updating changelogs, or running validation — covers bump.sh, validate.sh, content parity, and the release workflow
prefer-system-tools
Use when about to write a helper script — check if a native CLI tool can do it directly before reaching for Bun or Python
premortem
Use when the orchestrator must run a premortem before writing protocol.md — spawns N Haiku agents in parallel, each imagining one failure framing, then aggregates results into premortem.md
rnd-build
Run the Build phase for a specific task or wave from the existing RND plan.
rnd-building
Use when implementing code within the R&D pipeline — TDD discipline, pre-registration compliance, honest self-assessment, and verification artifact production
rnd-bump
Bump the plugin version (patch/minor/major) and add a CHANGELOG entry. Stages plugin.json and CHANGELOG.md, then offers to commit and optionally tag.
rnd-calibrate
Record a manual ground-truth verdict correction for a completed task. Reads the original verdict from the verification report and appends a correction record to calibration.jsonl.
rnd-calibration
Use when recording verdict data, detecting false verdicts, or injecting calibration stats into verifier prompts — enables the framework to learn from verification mistakes over time
rnd-cleanup
Use after a task passes verification — detect and remove dead code, orphan files, duplicate implementations, and stale comments introduced during the build, then re-verify and roll back if verification breaks
rnd-completion
Use after final SHIP verdict — guides branch management, PR creation, cleanup of .rnd/ artifacts, and development branch completion
rnd-data-science
Use when performing numerical analysis, financial calculations, data wiring, chart generation, or any analytical task requiring computation — Julia for statistics/charts/finance, DuckDB for SQL-expressible queries, CSV/Parquet aggregations, joins, and data exploration
rnd-debug-pipeline
Use when running the debug pipeline for a reported bug — defines the 4-phase flow, diagnosis report format, escalation criteria, and Builder handoff
rnd-debugging
Use when encountering any bug, test failure, or unexpected behavior within the R&D pipeline — systematic root cause analysis before proposing fixes
rnd-decomposition
Use when breaking a complex task into sub-tasks with pre-registration documents — structured hierarchical decomposition, dependency analysis, and testable success criteria
rnd-doc-polish
Use after SHIP verdict before committing — checks and updates CLAUDE.md, README.md, project docs, and stale inline comments to reflect what the pipeline just built
rnd-doctor
Check runtime environment readiness: CLI tools, hook scripts, artifact directory, plugin registration, version sync, and Julia MCP tools.
rnd-failure-modes
Use when verifying work or reviewing your own reasoning — a catalog of failure modes and anti-patterns that cause false PASSes, missed defects, and broken quality gates
rnd-formatting
Use before doc-polish and committing — detects the project's formatter and runs it on files changed by the pipeline
rnd-integration
Use when merging verified task outputs and running integration/system validation — ensures component-level PASS translates to system-level correctness
rnd-iteration
Use when handling build-verify feedback loops — receiving verification feedback, iteration budgets, escalation to re-planning
rnd-learning
Use when extracting or reading pipeline-discovered gotchas — defines when/how to capture learnings from iteration cycles and inject them into builder prompts
rnd-narrative
Generate a development narrative for a pipeline session — a prose story of what was built, key decisions, obstacles, insights, and what's left.
rnd-orchestration
Use when coordinating multi-agent R&D pipeline execution — provides pipeline overview, agent roles, information barriers, and gate criteria
rnd-plan
Run only the Planning phase: decompose a task into sub-tasks with pre-registration documents, dependency matrix, and execution schedule.
rnd-reality-auditing
Use when adversarially verifying external service contracts in Builder code — identifies SQL, API, MCP, SDK, and env-var assumptions, designs disproving experiments, and produces reality reports with VALID/INVALID/UNCHECKED verdicts backed by raw evidence
rnd-roadmapping
Use when planning multi-session work or managing roadmap milestones — defines roadmap.md format, milestone lifecycle, and how to update progress after SHIP
rnd-scaling
Use when deciding how much R&D pipeline ceremony a task needs — scales from trivial to high-stakes (dual verification)
rnd-verify
Run criticality-conditional verification on a built task. Verifies against pre-registered criteria with information barrier.
using-rnd-framework
Use when starting any conversation - establishes how to find and use R&D framework skills, requiring Skill tool invocation before ANY response
writing-skills
Use when creating new skills for the rnd-framework plugin — skill file format, frontmatter conventions, and best practices
rnd-bash-hook-testing
Use when writing or modifying tests for rnd-framework hook scripts — covers the test-helpers.sh framework, run_hook pattern, assertions, environment mocking, and test organization
rnd-bun-scripting
Use when writing helper scripts, utility tasks, or automation within the R&D pipeline — prefer Bun (TypeScript) over Python when available
rnd-code-review
Use when reviewing code changes for quality, correctness, and security — defines review categories, severity levels, verdicts, and the structured report format
rnd-committing
Use when creating git commits — enforces commit message style, length limits, and user confirmation before committing
rnd-fp-practices
Use alongside KISS practices to guide agents toward functional programming patterns — pure functions, data transformations, composition, command-query separation, and immutability
rnd-hook-authoring
Use when writing or modifying rnd-framework hook scripts — covers hook anatomy, exit code protocol, stdin parsing, fast-path patterns, and hooks.json registration
rnd-kiss-practices
Use when loading coding practices in Phase 0 discovery — language-specific KISS (Keep It Simple) rules to prevent over-engineering; detect project languages and read only the relevant `${CLAUDE_SKILL_DIR}/<lang>.md` files.
rnd-lib-sh-patterns
Use when writing hooks or lib scripts — covers lib.sh shared utilities, path predicates, response functions, stdin parsing, and the active_session_dir caching pattern
rnd-outside-view
Use in Phase 1 of rnd-start (after the premortem fan-out, before the Planner spawn) to inject a historical reference-class block — per-shape FAIL rates from the session corpus — into the Planner's context as a calibration anchor.
rnd-plugin-architecture
Use when working with the plugin structure — covers config dir detection, path matching, hooks.json, and hook event availability
rnd-plugin-versioning
Use when bumping plugin version, updating changelogs, or running validation — covers bump.sh, validate.sh, content parity, and the release workflow
rnd-prefer-system-tools
Use when about to write a helper script — check if a native CLI tool can do it directly before reaching for Bun or Python
rnd-premortem
Use when the orchestrator must run a premortem before writing protocol.md — spawns N Haiku agents in parallel, each imagining one failure framing, then aggregates results into premortem.md
rnd-using-rnd-framework
Use when starting any conversation - establishes how to find and use R&D framework skills, requiring Skill tool invocation before ANY response
rnd-writing-skills
Use when creating new skills for the rnd-framework plugin — skill file format, frontmatter conventions, and best practices
kiss-practices
Use when loading coding practices in Phase 0 discovery — language-specific KISS (Keep It Simple) rules to prevent over-engineering; detect project languages and read only the relevant `${CLAUDE_SKILL_DIR}/<lang>.md` files.
rnd-integrate
Run integration and system validation for a verified wave. Merges outputs, runs integration tests, checks for regressions.
rnd-scheduling
Use when planning execution waves from a dependency matrix — dependency-based scheduling and wave coordination
rnd-validate
Validate plugin structure: frontmatter, JSON files, hook references, and cross-references.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.