ai-research-explore
bylllllllama· 44 skills
Rigor Explore compatible skill slug for meaningful and potentially novel deep learning research candidates. Use when the researcher has chosen the task family, dataset, benchmark, evaluation method, provided SOTA references, and wants candidate-only exploration on top of current_research with auditable repo understanding, idea gating, fair comparison, and governed experiments written to explore_outputs/. Do not use for README-first trusted reproduction, open-ended direction finding, narrow code-only or run-only exploration, passive repo analysis, verified novelty claims, or implicit experimentation.
How to install ai-research-explore?
lllllllama/rigorpilot-skills/ai-research-explore
$npx -y skills add lllllllama/rigorpilot-skills --skill ai-research-exploreInstalls into the current project.
›Prefer a prompt? Paste this to your agent
Run `npx skills use "https://github.com/lllllllama/rigorpilot-skills" --skill "lllllllama/rigorpilot-skills/ai-research-explore"` and follow the generated skill instructions now. Read its complete output, redirecting it to a temporary file first if necessary. Resolve relative paths from the supporting-files directory it provides.
Use the skills in "https://github.com/lllllllama/rigorpilot-skills" that are relevant to the current task. Run `npx skills add "https://github.com/lllllllama/rigorpilot-skills"` and select the relevant skills, then follow their instructions.
SKILL.md
| 1 | # ai-research-explore |
| 2 | |
| 3 | ## Purpose |
| 4 | |
| 5 | Use this as the Rigor Explore compatible skill slug after the researcher |
| 6 | explicitly authorizes candidate-only work on top of a durable |
| 7 | `current_research` anchor. The installed slug remains `ai-research-explore` for |
| 8 | compatibility. Rigor Explore is for meaningful and potentially novel deep |
| 9 | learning research candidates while preserving scientific rigor, comparability, |
| 10 | reproducibility, and auditable collaboration. Novelty and significance remain |
| 11 | hypotheses before literature contrast, ablation evidence, and fair comparison. |
| 12 | The skill does not promise autonomous discovery, global benchmark completeness, |
| 13 | novelty proof, or trusted reproduction success. |
| 14 | |
| 15 | Start from the shared operating principles in |
| 16 | `../../references/agent-operating-principles.md`, then load |
| 17 | `../../references/research-rigor-principles.md` for research claims and |
| 18 | `../../references/deep-learning-experiment-principles.md` when experiment |
| 19 | details affect comparability or reproducibility. |
| 20 | |
| 21 | ## Fit |
| 22 | |
| 23 | Use this skill only when the request has both: |
| 24 | |
| 25 | - Explicit exploration authorization such as candidate-only work, isolated |
| 26 | branch or worktree, sweep, several variants, or exploratory ranking. |
| 27 | - A durable `current_research` context such as a branch, commit, checkpoint, |
| 28 | run record, or already-trained local model state. |
| 29 | |
| 30 | Keep narrow code-only requests on `explore-code`. Keep narrow run-only requests |
| 31 | on `explore-run`. Keep passive repository analysis on `analyze-project`. Keep |
| 32 | README-first reproduction on `ai-research-reproduction`. |
| 33 | |
| 34 | ## Research Rhythm |
| 35 | |
| 36 | Use a two-loop rhythm: |
| 37 | |
| 38 | - Outer loop: understand the repository, freeze task/dataset/evaluation/budget, |
| 39 | preserve user ideas, map sources, gate ideas, and decide whether the next |
| 40 | experiment is worth running. |
| 41 | - Inner loop: make one bounded candidate change or run, smoke-check it, collect |
| 42 | evidence, rank it against the current anchor, and either stop or return to the |
| 43 | outer loop with the new evidence. |
| 44 | |
| 45 | This rhythm is a guide, not a rigid autonomous loop. Stop at explicit blockers, |
| 46 | unclear scientific meaning, exhausted budget, missing anchor/evaluation, or a |
| 47 | human checkpoint. |
| 48 | |
| 49 | ## Workflow |
| 50 | |
| 51 | 1. Confirm `current_research` and explicit explore-lane authorization. |
| 52 | 2. Accept either legacy `variant_spec` or higher-level `research_campaign`. |
| 53 | 3. In campaign mode, freeze the task, dataset, benchmark, evaluation source, |
| 54 | SOTA reference, and budget before candidate work. |
| 55 | 4. Build only the repo-understanding artifacts needed for the current campaign, |
| 56 | usually through `analyze-project`. |
| 57 | 5. Run bounded, cache-first source lookup when source support matters; prefer |
| 58 | local curated literature such as Zotero if available, then seed sources, |
| 59 | repo-local locators, public locators, or optional web lookup. Treat lookup as |
| 60 | source resolution, not an open-ended literature search. |
| 61 | 6. Preserve researcher-provided ideas, optionally add a small bounded set of |
| 62 | single-variable seed ideas, and rank ideas with explicit gates and score |
| 63 | breakdowns. |
| 64 | 7. Prefer one clear candidate at a time. Use `explore-code` for bounded code |
| 65 | adaptation and `explore-run` for short-cycle trials or sweeps. |
| 66 | 8. Use `minimal-run-and-audit` or `run-train` only when the exploratory plan |
| 67 | requires real execution evidence. |
| 68 | 9. Write candidate-only outputs to `analysis_outputs/`, `sources/`, and |
| 69 | `explore_outputs/` as appropriate; never present exploratory gains as trusted |
| 70 | reproduction success. Include `SCIENTIFIC_CHANGELOG.md` and |
| 71 | `COMPARABILITY_REPORT.md` for candidate scientific meaning and comparison |
| 72 | boundaries. |
| 73 | |
| 74 | ## Ranking and Evidence |
| 75 | |
| 76 | - Before execution, prioritize candidates by expected gain, cost, success |
| 77 | likelihood, patch surface, dependency drag, evaluation risk, and rollback |
| 78 | ease. |
| 79 | - After execution, rank by real evidence first: command status, observed |
| 80 | metrics, artifacts, changed paths, smoke results, and reproducibility notes. |
| 81 | - Keep researcher-provided `evaluation_source` and `sota_reference` frozen for |
| 82 | the campaign; do not claim they are globally complete. |
| 83 | - If the top ideas are too close or the implementation cannot be decomposed into |
| 84 | auditable units, stop for a checkpoint instead of silently choosing. |
| 85 | |
| 86 | ## Campaign Inputs |
| 87 | |
| 88 | `research_campaign` is preferred for Rigor Explore campaigns, but it should |
| 89 | stay minimal. The durable core is: |
| 90 | |
| 91 | - `current_research` |
| 92 | - `task_family` |
| 93 | - `dataset` |
| 94 | - `benchmark` |
| 95 | - `evaluation_source` |
| 96 | - `sota_reference` |
| 97 | - `compute_budget` |
| 98 | |
| 99 | Use `candidate_ideas`, `variant_spec`, `research_lookup`, `idea_policy`, |
| 100 | `idea_generation`, `source_constraints`, `feasibility_policy`, `baseline_gate`, |
| 101 | and `execution_policy` as optional guidance, not as fields the agent must fill |
| 102 | for every campaign. See `references/research-campaign-spec.md` for the advanced |
| 103 | schema and artifact expectations. |
| 104 | |
| 105 | ## Reference Loading |
| 106 | |
| 107 | - Load `references/ai-research-explore-policy.md` for lane safety and candidate |
| 108 | semantics. |
| 109 | - Load `references/research-campaign-spec.md` only when a campaign file is |
| 110 | present or the user asks for Rigor Explore campaign governance. |
| 111 | - Load `../../references/explore-variant-spec.md` for run-level variant matrix |
| 112 | details. |
| 113 | - Load `../../references/research-rigor-principles.md` before making novelty, |
| 114 | contribution, SOTA, or comparability statements. |
| 115 | - Load `../../references/deep-learning-experiment-principles.md` when training, |
| 116 | evaluation, baseline, ablation, metric, checkpoint, or dataset details matter. |
| 117 | - Use `scripts/orchestrate_explore.py` and `scripts/write_outputs.py` for the |
| 118 | existing deterministic artifact workflow. |
Related
6 picksType
- azure-aiUse for Azure AI: Search, Speech, OpenAI, Document Intelligence. Helps with search, vector/hybrid search, speech-to-text, text-to-speech, transcription, OCR.Jul 2026
- ai-research-reproductionRigor Reproduce compatible skill slug for README-first deep learning repository reproduction.Jul 2026
- explore-codeRigor Improve implementation leaf skill for auditable candidate implementation in deep learning research repositories.Jul 2026
- run-trainRigor Train skill for deep learning research repositories. Use when a documented or selected training command should be run conservatively for startup…Jul 2026
- explore-runRigor Improve / Rigor Explore run leaf skill for bounded exploratory evidence in deep learning research repositories.Jul 2026
- minimal-run-and-auditRigor Run skill for README-first deep learning repo reproduction.Jul 2026
