What question does this Skill answer?
Start with the target repository's README and existing files. If they do not settle a concrete detail that could change whether a result is comparable—such as a dataset split, preprocessing step, evaluation protocol, checkpoint mapping, or runtime assumption—this helper can look for the narrow missing evidence in primary paper sources.
Its output is intentionally small: the relevant sources, a reproduction-focused answer, any README–paper conflict, and a clear boundary between direct evidence and inference. A plausible guess is not a substitute for missing evidence.
| Good fit | Not its job |
|---|---|
| Verify a dataset version or split that changes comparability. | Summarize an entire paper for general understanding. |
| Check preprocessing, evaluation, checkpoint, or runtime assumptions against primary sources. | Scan a repository, set up an environment, or run commands. |
| Record a conflict between the paper and repository instructions. | Silently replace the README or infer missing protocol details as facts. |
How it fits the RigorPilot workflow
The upstream RigorPilot README describes `paper-context-resolver` as one of two optional orchestration helpers. It is helper-tier and normally invoked to resolve a narrow evidence gap inside a README-first reproduction flow; it is not the main reproduction workflow or an experiment runner.
The repository documents a single-Skill install command for its main reproduction Skill and `--all` for the companion and leaf Skills. So the command below installs the full collection. If you only need this helper, open the exact upstream folder and follow your agent's supported manual project-Skill path instead of assuming a single-Skill installer command is documented.
npx skills add llllllllama/RigorPilot-Skills --all- Provide a target repository and one concrete reproduction-critical question.
- Keep README and repository evidence as the starting point; cite primary paper sources for missing details.
- Separate source-backed facts from inference and stop if the evidence does not support a trustworthy answer.
Limits and evidence status
The current upstream Skill says not to use it for general paper summaries, title-only lookups, broad repository scans, environment setup, or command execution. It supplements the main reproduction flow and does not automatically establish that an experiment is reproducible.
SkillSignal reviewed the published README and SKILL.md for this guide. We did not run the Skill, compare model outputs, or independently reproduce a paper result; this page is a source-based explanation, not a benchmark or scientific certification.
Primary sources
Check the current upstream files before installing because repository contents and instructions can change.