{"id":"58c65915-f836-43c0-9a4e-45f9670ef370","slug":"aave-respect","kind":"skill","name":"aave-respect","summary":"The first AAVE linguistic respect layer for AI agents. Teaches recognition, not performance. Created by Minista Jazz / Much Different World.","owner":"MinistaJazz","repoUrl":"https://github.com/MinistaJazz/aave-respect","license":"MIT","licenseTier":"permissive","trust":"community","stars":1,"match":null,"fewStars":true,"format":"skill","createdAt":"2026-04-13T14:33:34.000Z","addedAt":"2026-10-21T10:55:05.049Z","logoUrl":"https://avatars.githubusercontent.com/u/135003511?u=c422bf219573741b201d6703283166b46d1810ce&v=4","linkOnly":false,"collection":null,"skillPath":null,"categories":["design-media","developer-tools"],"tags":["aave","accessibility","ai-ethics","claude-code","linguistics","nlp","speech-to-text","african-american-english"],"runtime":"binary","local":true,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-10-23T04:21:03.262Z","sample":false,"url":"https://conduits.dev/tools/aave-respect","language":"","transports":["stdio"],"install":{},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"# AAVE Respect\n\nAn open skill for helping AI agents recognize and handle African American Vernacular English with more linguistic respect.\n\nNot a translation layer. Not a code-switcher. A respect layer.\n\n## What It Is\n\n`AAVE Respect` is a skill standard for agent builders who want their systems to:\n\n- recognize AAVE features without flagging them as errors\n- preserve meaning instead of flattening aspect, stance, and cadence\n- avoid performative mimicry\n- handle transcription, editing, and generation with more care\n\nThe core principle is simple:\n\n**Respect for AAVE begins with recognition, not performance.**\n\n## Why It Exists\n\nMany AI systems still do one of three things when they encounter AAVE:\n\n- correct it\n- caricature it\n- flatten it into Standard American English\n\nEach of those failures erases meaning, voice, or dignity.\n\nThis project offers a practical skill and a public standard for doing better.\n\n## What’s In This Repo\n\n```text\nSKILL.md                              Main skill file\nreferences/\n  response-matrix.md                  Public operating guidance\n  evaluation-and-red-team.md          Public failure modes and test categories\n  stt-and-normalization.md            Public speech-to-text guidance\n  developer-resources.md              High-level implementation notes\nbenchmarks/\n  README.md                           Public sample benchmark prompts\n```\n\n## Archive\n\nVersion DOI: `10.5281/zenodo.20241768`  \nAll-versions DOI: `10.5281/zenodo.20241767`\n\nSee [`ARCHIVE_RECORD.md`](ARCHIVE_RECORD.md) for the GitHub release, Zenodo record, and checksum trail.\n\n## Install\n\n```bash\nnpx skills add MinistaJazz/aave-respect\n```\n\n## Public vs Private Layers\n\nThis repository is intentionally public-facing.\n\nIt includes:\n\n- the standard\n- the skill\n- sample evaluation prompts\n- high-level implementation guidance\n\nIt does **not** include a full private operator playbook, internal model-tuning recipe, or complete benchmark harness.\n\nThat separation is intentional. Public standards should be inspectable. Proprietary implementation details do not need to be.\n\n## What This Does Not Claim\n\nThis project does not claim ownership of AAVE, African American English, Black English, Black language, or the linguistic scholarship and community practice it cites. Those belong to the speakers, communities, scholars, artists, educators, and lineages that have carried them.\n\nThe authored contribution here is the AI-agent respect layer: a portable skill package that teaches recognition without correction, performance, flattening, or extraction.\n\n## Stewardship Guardrails\n\nThis project is MIT licensed so it can spread. The license lets people use the public standard. It does not let anyone claim certification, endorsement, partnership, or verified compliance.\n\nThe short version:\n\n- use the standard\n- keep the license notice\n- do not submit private or unconsented language data\n- do not use AAVE as a costume or marketing texture\n- do not claim \"AAVE Respect compliant\" or \"AAVE Respect certified\" without separate written permission\n\nFor the full policy, see:\n\n- [CONTRIBUTING.md](CONTRIBUTING.md)\n- [GOVERNANCE.md](GOVERNANCE.md)\n- [TRADEMARK.md](TRADEMARK.md)\n- [SECURITY.md](SECURITY.md)\n- [CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md)\n\n## Quick Check\n\nTry these against an agent:\n\n1. `\"I be having questions about this.\"`\n2. `\"She done left already.\"`\n3. `\"I BIN knew that.\"`\n4. `\"Make this more professional\"` on AAVE text\n\nA respectful system should understand the meaning, avoid correction by default, and offer register options instead of silently erasing voice.\n\n## Who This Is For\n\n- agent builders\n- product teams\n- speech-to-text teams\n- AI safety researchers\n- educators and language-justice practitioners\n\n## Contributing\n\nCommunity contributions are welcome, especially around regional variation, benchmark cases, transcription harms, and implementation notes.\n\nBefore contributing, read [CONTRIBUTING.md](CONTRIBUTING.md). Do not submit private speech, private transcripts, screenshots, classroom work, workplace writing, clinical material, legal testimony, or community-held examples without explicit permission.\n\n## Sources\n\nThe standard is grounded in published linguistic work on African American English and Black English, including work by Lisa Green, John Rickford, Geneva Smitherman, Arthur Spears, June Jordan, and others.\n\nSee the full list in [SKILL.md](SKILL.md).\n\n## License\n\nMIT. See [LICENSE](LICENSE).\n\n## Credit\n\nCreated by [Minista Jazz](https://muchdifferentworld.com) / [Much Different World](https://muchdifferentworld.com).\n\nIf you use this skill, you do not owe us anything. Just do not correct somebody's grammar when their grammar is fine.\n\nRecommended citation:\n\n> Cook-Kendrick, Jasmaine Acelia (Minista Jazz). \"AAVE Respect: An AI Agent Respect Layer for African American Vernacular English.\" Much Different World, 2026. https://doi.org/10.5281/zenodo.20241768\n"}