{"id":51575008,"url":"https://github.com/tjhavranek/erc-ai-feedback","last_synced_at":"2026-07-10T23:30:54.851Z","repository":{"id":360663132,"uuid":"1251103176","full_name":"tjhavranek/erc-ai-feedback","owner":"tjhavranek","description":"Prompt + rubric for AI triage of ERC Starting/Consolidator Grant drafts before peer review. Doesn't replace human reviewers — frees workshop time for what AI can't address.","archived":false,"fork":false,"pushed_at":"2026-06-24T12:27:45.000Z","size":266,"stargazers_count":4,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-24T14:12:51.948Z","etag":null,"topics":["ai-tools","claude-code","codex","erc","european-research-council","grant-writing","peer-review","research-support"],"latest_commit_sha":null,"homepage":null,"language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/tjhavranek.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":"CITATION.cff","codeowners":null,"security":null,"support":null,"governance":null,"roadmap":"docs/roadmap.md","authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-05-27T08:43:52.000Z","updated_at":"2026-06-24T12:27:57.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/tjhavranek/erc-ai-feedback","commit_stats":null,"previous_names":["tjhavranek/erc-ai-feedback"],"tags_count":5,"template":false,"template_full_name":null,"purl":"pkg:github/tjhavranek/erc-ai-feedback","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tjhavranek%2Ferc-ai-feedback","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tjhavranek%2Ferc-ai-feedback/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tjhavranek%2Ferc-ai-feedback/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tjhavranek%2Ferc-ai-feedback/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/tjhavranek","download_url":"https://codeload.github.com/tjhavranek/erc-ai-feedback/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tjhavranek%2Ferc-ai-feedback/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":35346644,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-07-10T02:00:06.465Z","response_time":60,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai-tools","claude-code","codex","erc","european-research-council","grant-writing","peer-review","research-support"],"created_at":"2026-07-10T23:30:53.056Z","updated_at":"2026-07-10T23:30:54.844Z","avatar_url":"https://github.com/tjhavranek.png","language":null,"funding_links":[],"categories":[],"sub_categories":[],"readme":"# erc-ai-feedback\n\nThis is a small package for ERC Starting and Consolidator Grant\napplicants. It provides a rubric-based pre-review of a draft\nproposal that the applicant runs in one chat session with a current\nfrontier model. The intent is to clear routine structural problems\n— vague hypotheses, hidden incrementalism, missing risk logic,\nweak panel fit, CV–project disconnect — before workshop time is\nspent on them. It does not replace human review. It is for\napplicants and applicant-authorised workshop preparation only: do\nnot use it on a proposal you are evaluating as an ERC panel member\nor remote reviewer, since the ERC [prohibits reviewers from\nuploading proposals under evaluation to external AI\nsystems](https://erc.europa.eu/news-events/news/erc-clarifies-limits-ai-use-grant-evaluation).\n\nTwo files do the work: the locked evaluation rubric in\n[`shared/rubric_locked.md`](shared/rubric_locked.md) and the prompt\nin [`basic/prompt.md`](basic/prompt.md). The rubric contains the\nverbatim Annex 1 Criterion 1 and Criterion 2 sub-questions from the\nERC *Guide for Peer Reviewers* for Starting and Consolidator Grant\ncalls, plus source-checked notes on scoring scales, proposal\nstructure, eligibility windows, panel structure, and the ERC's\nAI-use rules. The prompt loads the rubric, applies structural caps\nthat fire when an expected element is missing, and produces an\nevidence-grounded critique.\n\nThis package is the ERC-specific application of two general\nresearch-audit projects by the same author group:\n[`tjhavranek/research-audit-duel-protocol`](https://github.com/tjhavranek/research-audit-duel-protocol)\n(manual multi-model adversarial-debate protocols) and\n[`tjhavranek/mad-research`](https://github.com/tjhavranek/mad-research)\n(Claude Code skills that automate the same audit discipline for\nClaude + Codex). The Advanced pre-review here is a direct\nextension of the latter; the multi-model convergence walkthrough\nin [`docs/multi_model_walkthrough.md`](docs/multi_model_walkthrough.md)\nis a direct application of the former.\n\nA sibling project,\n[`tjhavranek/paper-workshop`](https://github.com/tjhavranek/paper-workshop),\nis a heavier Claude Code skill that runs an adversarial expert\nworkshop on a full manuscript and then rebuilds it; it is built for\npapers but can be adapted to proposals. It is noted here for\ncompleteness rather than recommended for ERC use, because it does\nsomething this pre-review deliberately avoids: it proposes and writes\nedits, whereas this package only identifies problems and leaves the\ndrafting to the applicant. It is not calibrated to the ERC rubric and\nis not part of this package.\n\n[![License](https://img.shields.io/badge/license-MIT-green)](LICENSE)\n[![DOI](https://zenodo.org/badge/1251103176.svg)](https://doi.org/10.5281/zenodo.20829165)\n\n## Use\n\n1. Open a paid chat session with a current frontier model. Confirm\n   training opt-out is on. (Provider settings change; see Privacy\n   below.)\n2. Copy the text between the `=== PROMPT BEGIN ===` and `=== PROMPT\n   END ===` markers in [`basic/prompt.md`](basic/prompt.md). Paste\n   it into the chat.\n3. Paste your current draft after the prompt. State which stage you\n   are at:\n   - **Pitch** — idea sketch, or abstract plus draft CV.\n   - **Part I + CV** — Extended Synopsis (Part I of the Scientific\n     Proposal) in progress, with a draft CV.\n   - **Full proposal** — complete B1 (cover-page abstract, Part I,\n     CV and track record), plus any available B2 draft (Part II;\n     required for final ERC submission).\n4. Read the output. The model returns the top findings ranked by\n   severity, each grounded in a verbatim quote from your draft;\n   which structural caps fired; a short list of what the next draft\n   stage needs to add; and a paragraph you can attach when sending\n   the draft to a workshop reader.\n\n## Privacy\n\nUse a paid tier with training opt-out enabled, or remove names and\ninstitutions before pasting a proposal draft into any chat product.\nProvider UI labels change every few months, so check the provider's\ncurrent data-controls page rather than relying on setting names\ncopied here:\n[ChatGPT data controls](https://help.openai.com/en/articles/7730893-chatgpt-privacy-practices),\n[Anthropic Claude privacy](https://privacy.claude.com/en/articles/10023580-is-my-data-used-for-model-training),\nand [Gemini activity and privacy](https://support.google.com/gemini/answer/13594961).\nPasting an unpublished draft into any chat product carries residual\nrisk even when training is off. If the proposal contains embargoed,\npatent-sensitive, or otherwise non-public material, sanitise it or\ndo not use the tool. If someone other than the applicant runs the\npre-review, follow the consent rule in\n[`docs/for_workshop_leaders.md`](docs/for_workshop_leaders.md).\n\n## Limitations\n\nThe tool is calibrated for ERC Starting and Consolidator Grants\nonly — not Advanced, Synergy, Proof of Concept, or ERC Plus. It\nis in English; the ERC working language of evaluation is English.\nIt catches routine structural problems and does not judge novelty\nin a specific field, panel strategy, or interview credibility —\nthese remain workshop, mentor, and research-office work. Frontier\nmodels confidently miss field-specific failure modes; the\napplicant should push back on findings they disagree with on\nmerit.\n\nThe ERC rubric changes annually. The version in\n[`shared/rubric_locked.md`](shared/rubric_locked.md) was last\nverified on 2026-05-30 against ERC Work Programme 2026 and the\n*Guide for Peer Reviewers* v6.0 (30 September 2025). If you read\nthis twelve months later, re-verify before use.\n\n## For workshop leaders\n\nIf you are running an ERC workshop and want to recommend the tool\nto applicants, the operational note at\n[`docs/for_workshop_leaders.md`](docs/for_workshop_leaders.md)\ncovers the hand-off template and the consent rule that applies\nwhen a reader runs the tool on someone else's draft.\n\n## Other ways to use the package\n\nThe single-prompt pre-review above is what most applicants will\nuse. The repository also contains:\n\n- A three-prompt sequence in [`standard/`](standard/) for\n  applicants who want a more thorough single-model pass\n  (Panel Reviewer, then Devil's Advocate, then Writing Coach).\n- A Claude Code + Codex CLI integration in\n  [`advanced/`](advanced/) that runs the\n  [`mad-research`](https://github.com/tjhavranek/mad-research)\n  protocol against the same ERC rubric (experimental; not yet\n  end-to-end pilot-verified — see\n  [`advanced/README.md`](advanced/README.md)).\n- A question-bank and answer stress-test prompt pair in\n  [`mock_interview/`](mock_interview/) for Step-2 applicants\n  preparing for the panel interview.\n- A resubmission audit prompt in [`resubmission/`](resubmission/)\n  that checks whether a new draft engages with a previous\n  Evaluation Report.\n\nSynthetic test fixtures used to exercise the prompts are in\n[`tests/fixtures/`](tests/fixtures/). Roadmap and development\nnotes are in [`docs/roadmap.md`](docs/roadmap.md).\n\n## Citation, licence, maintainer\n\nDeveloped by Tomáš Havránek and Zuzana Iršová, Charles University,\nPrague. MIT licence; see\n[`LICENSE`](LICENSE). The ERC materials cited in the rubric remain\nthe property of the European Research Council Executive Agency\nand are referenced under fair use with source URLs in\n[`shared/rubric_locked.md`](shared/rubric_locked.md). Archived on\nZenodo with a DOI that resolves to the latest release:\n[`https://doi.org/10.5281/zenodo.20829165`](https://doi.org/10.5281/zenodo.20829165).\nFor citation metadata see [`CITATION.cff`](CITATION.cff).\n\nBug reports and factual corrections to the rubric: GitHub Issues.\nPlease do not paste real proposal content in public issues.\nERC-panel-veteran feedback on the prompt is particularly welcome.\n\nRoadmap and development notes are in\n[`docs/roadmap.md`](docs/roadmap.md). The audit history of each\nversion is in [`CHANGELOG.md`](CHANGELOG.md).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftjhavranek%2Ferc-ai-feedback","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ftjhavranek%2Ferc-ai-feedback","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftjhavranek%2Ferc-ai-feedback/lists"}