[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M79BAYE0DEVKPVFSVSS5CW7W":3,"skill-same:01M79BAYE0DEVKPVFSVSS5CW7W":54},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":28,"commitSha":29,"skillDirectory":30,"forksCount":13,"repositoryHtmlUrl":31,"repositoryHomepageUrl":6,"ownerHtmlUrl":32,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":33,"install":34,"resources":47},"01M79BAYE0DEVKPVFSVSS5CW7W","agmsg-orchestration",null,"Coordinate structured agmsg task orchestration between a Claude Code orchestrator and Codex workers. Use when Codex or Claude Code needs to run or supervise AGMSG-TASK \u002F AGMSG-RESULT \u002F AGMSG-ACCEPTANCE workflows, bootstrap workers in herdr panes, manage .orchestration artifacts, act as an agmsg worker, or document the flue-pi style orchestration protocol without installing the Hermes Agents runtime.","mryfmo\u002Fherness-self-improvement","mryfmo","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F11512262?v=4","01M71K2RDZFFQKYGMHMS5Y9PW1","01M79BAQKWWXZCC4B5C7M29WTS",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fmryfmo\u002Fherness-self-improvement\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fagmsg-orchestration\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: agmsg-orchestration\ndescription: Coordinate structured agmsg task orchestration between a Claude Code orchestrator and Codex workers. Use when Codex or Claude Code needs to run or supervise AGMSG-TASK \u002F AGMSG-RESULT \u002F AGMSG-ACCEPTANCE workflows, bootstrap workers in herdr panes, manage .orchestration artifacts, act as an agmsg worker, or document the flue-pi style orchestration protocol without installing the Hermes Agents runtime.\nmetadata:\n  scope: org\n  owner: mryfmo\n  last-verified: 2026-07-21\n  freshness: 90d\n  provenance: repo=mryfmo\u002Fdotfiles; license=MIT\n---\n\n# agmsg orchestration\n\nUse this skill for structured multi-agent work where a Claude Code orchestrator assigns bounded tasks to Codex workers through `agmsg` teams. Use the regular `agmsg` skill for simple send\u002Finbox\u002Fhistory commands.\n\n## Architecture\n\n- Claude Code is the orchestrator: it writes task files, starts workers, reviews artifacts, and sends acceptance or revision messages.\n- Codex workers execute one assigned task: they read the task file, obey file and action constraints, write artifacts, and send the required result message.\n- `agmsg` is the message bus. Use only scripts under `~\u002F.agents\u002Fskills\u002Fagmsg\u002Fscripts\u002F`.\n- `herdr` panes are optional worker terminals; they are a launch surface, not the protocol.\n- This skill adopts only the Hermes Skill Subset ideas: `SKILL.md` structure, progressive disclosure, activation metadata, task\u002Ferror\u002Fuser-correction skill decisions, and separated candidate\u002Fpromoted\u002Frejected\u002Fmerged registries. Do not introduce Hermes Agents runtime, memory, profiles, personalities, toolsets, plugins, UI, or automation framework.\n\n## Message Contract v1\n\nSend messages as single-line records so inbox\u002Fhistory output stays parseable.\n\n`AGMSG-TASK v1` fields:\n\n```text\nAGMSG-TASK v1 task_id=\u003Cid> repo=\u003Cabsolute-repo-path> task_file=\u003Cpath>\nallowed_files=\u003Cpaths-or-see-task-file-section> forbidden_actions=\u003Csemicolon-list>\nexpected_result_file=\u003Cpath> expected_validation_file=\u003Cpath>\nexpected_sandbox_file=\u003Cpath> expected_learning_file=\u003Cpath>\nexpected_autoskill_file=\u003Cpath> done_signal=AGMSG-RESULT max_turns=\u003Cn>\nnote=act-as-worker-\u003Ctask-or-role>\n```\n\n`AGMSG-RESULT v1` fields:\n\n```text\nAGMSG-RESULT v1 task_id=\u003Cid> status=ready_for_review|blocked\nreport=\u003Cpath> validation=\u003Cpath> sandbox=\u003Cpath> learning=\u003Cpath> autoskill=\u003Cpath>\n```\n\n`AGMSG-ACCEPTANCE v1` fields:\n\n```text\nAGMSG-ACCEPTANCE v1 task_id=\u003Cid> status=accepted|revise reason=\u003Cshort-reason> next_action=\u003Caction>\n```\n\nLiveness messages:\n\n```text\nAGMSG-PING v1 task_id=\u003Cid> reason=\u003Cshort-reason>\nAGMSG-PONG v1 task_id=\u003Cid> status=alive|blocked note=\u003Cshort-note>\n```\n\n## `.orchestration` Workspace Layout\n\n- `tasks\u002F`: orchestrator-authored task specs.\n- `reports\u002F`: worker reports and blocked-task reports.\n- `validation\u002F`: command output and validation evidence.\n- `acceptance\u002F`: orchestrator acceptance, revision, or rejection records.\n- `sandboxes\u002F`: OpenSandbox records or documented fallback records.\n- `autoskill\u002Fconfig\u002F`, `autoskill\u002Finputs\u002F`, `autoskill\u002Fruns\u002F`, `autoskill\u002Foutputs\u002F`: redacted AutoSkill artifacts.\n- `learning\u002F`: task learning triage records.\n- `learning\u002Frule_candidates\u002F`: candidate reusable rules only.\n- `skills\u002Fcandidates\u002F`, `skills\u002Fpromoted\u002F`, `skills\u002Frejected\u002F`, `skills\u002Fmerged\u002F`: separated skill registry states.\n- `agmsg\u002F`: exported or summarized agmsg history when needed for review.\n\n## Orchestrator Playbook\n\n1. Join or confirm the agmsg team and identities with the `agmsg` scripts.\n2. Create the `.orchestration` directories before assigning work.\n3. Write a task file that includes objective, scope, allowed files, forbidden actions, expected artifacts, validation commands, and max turns.\n4. Start worker panes if needed. With herdr, send the worker prompt text into the pane and submit it with a carriage return; sending text alone may leave it unsubmitted.\n5. Configure delivery deliberately. `delivery.sh set turn` is useful for turn-end inbox checks; changing delivery mode can kill project watcher processes, so do it before starting long-running project watchers.\n6. Send `AGMSG-TASK v1` with the exact artifact paths and `done_signal=AGMSG-RESULT`.\n7. Track `max_turns`. Use `AGMSG-PING` for liveness if a worker stalls.\n8. On `AGMSG-RESULT`, read the task file and every referenced artifact before deciding.\n9. Send `AGMSG-ACCEPTANCE v1 status=accepted` when done, or `status=revise` with a narrow `reason` and `next_action` when more work is required.\n\n## Worker Playbook\n\n1. Read the full `AGMSG-TASK v1` message.\n2. Switch to the `repo` and read `task_file` before editing or running validations.\n3. Treat `allowed_files` as the edit boundary. If it says to see the task file, read that section and follow it exactly.\n4. Do not perform any `forbidden_actions`.\n5. Write artifacts to the exact expected paths. Do not invent alternate paths.\n6. Put command outputs and validation evidence in `expected_validation_file`.\n7. Put sandbox\u002FOpenSandbox status or fallback rationale in `expected_sandbox_file`.\n8. Put reusable learning triage in `expected_learning_file`; do not promote rules directly unless the task explicitly allows it.\n9. Put AutoSkill run status or a not-used record in `expected_autoskill_file`.\n10. If blocked, still write the report and evidence paths that explain the blocker.\n11. Reply with the requested `done_signal`, normally `AGMSG-RESULT v1`, and include all artifact paths.\n\n## Pitfalls\n\n- Do not start work from the agmsg message alone; read `task_file` first.\n- Do not edit outside `allowed_files`, even for convenient cleanup.\n- Do not perform forbidden actions such as dependency changes, gate changes, product changes, promotion decisions, image builds, or LLM calls when listed.\n- Do not collapse candidate, promoted, rejected, and merged skill registry states into one directory.\n- Do not put secrets, raw logs with credentials, or unredacted AutoSkill inputs in artifacts.\n- Do not install Hermes Agents runtime for this protocol.\n- Do not assume `herdr` text injection submits automatically; send Enter\u002FCR.\n- Do not treat `AGMSG-ACCEPTANCE status=revise` as a new task unless it changes the task file or explicitly provides a next action.\n","# agmsg orchestration\n\nUse this skill for structured multi-agent work where a Claude Code orchestrator assigns bounded tasks to Codex workers through `agmsg` teams. Use the regular `agmsg` skill for simple send\u002Finbox\u002Fhistory commands.\n\n## Architecture\n\n- Claude Code is the orchestrator: it writes task files, starts workers, reviews artifacts, and sends acceptance or revision messages.\n- Codex workers execute one assigned task: they read the task file, obey file and action constraints, write artifacts, and send the required result message.\n- `agmsg` is the message bus. Use only scripts under `~\u002F.agents\u002Fskills\u002Fagmsg\u002Fscripts\u002F`.\n- `herdr` panes are optional worker terminals; they are a launch surface, not the protocol.\n- This skill adopts only the Hermes Skill Subset ideas: `SKILL.md` structure, progressive disclosure, activation metadata, task\u002Ferror\u002Fuser-correction skill decisions, and separated candidate\u002Fpromoted\u002Frejected\u002Fmerged registries. Do not introduce Hermes Agents runtime, memory, profiles, personalities, toolsets, plugins, UI, or automation framework.\n\n## Message Contract v1\n\nSend messages as single-line records so inbox\u002Fhistory output stays parseable.\n\n`AGMSG-TASK v1` fields:\n\n```text\nAGMSG-TASK v1 task_id=\u003Cid> repo=\u003Cabsolute-repo-path> task_file=\u003Cpath>\nallowed_files=\u003Cpaths-or-see-task-file-section> forbidden_actions=\u003Csemicolon-list>\nexpected_result_file=\u003Cpath> expected_validation_file=\u003Cpath>\nexpected_sandbox_file=\u003Cpath> expected_learning_file=\u003Cpath>\nexpected_autoskill_file=\u003Cpath> done_signal=AGMSG-RESULT max_turns=\u003Cn>\nnote=act-as-worker-\u003Ctask-or-role>\n```\n\n`AGMSG-RESULT v1` fields:\n\n```text\nAGMSG-RESULT v1 task_id=\u003Cid> status=ready_for_review|blocked\nreport=\u003Cpath> validation=\u003Cpath> sandbox=\u003Cpath> learning=\u003Cpath> autoskill=\u003Cpath>\n```\n\n`AGMSG-ACCEPTANCE v1` fields:\n\n```text\nAGMSG-ACCEPTANCE v1 task_id=\u003Cid> status=accepted|revise reason=\u003Cshort-reason> next_action=\u003Caction>\n```\n\nLiveness messages:\n\n```text\nAGMSG-PING v1 task_id=\u003Cid> reason=\u003Cshort-reason>\nAGMSG-PONG v1 task_id=\u003Cid> status=alive|blocked note=\u003Cshort-note>\n```\n\n## `.orchestration` Workspace Layout\n\n- `tasks\u002F`: orchestrator-authored task specs.\n- `reports\u002F`: worker reports and blocked-task reports.\n- `validation\u002F`: command output and validation evidence.\n- `acceptance\u002F`: orchestrator acceptance, revision, or rejection records.\n- `sandboxes\u002F`: OpenSandbox records or documented fallback records.\n- `autoskill\u002Fconfig\u002F`, `autoskill\u002Finputs\u002F`, `autoskill\u002Fruns\u002F`, `autoskill\u002Foutputs\u002F`: redacted AutoSkill artifacts.\n- `learning\u002F`: task learning triage records.\n- `learning\u002Frule_candidates\u002F`: candidate reusable rules only.\n- `skills\u002Fcandidates\u002F`, `skills\u002Fpromoted\u002F`, `skills\u002Frejected\u002F`, `skills\u002Fmerged\u002F`: separated skill registry states.\n- `agmsg\u002F`: exported or summarized agmsg history when needed for review.\n\n## Orchestrator Playbook\n\n1. Join or confirm the agmsg team and identities with the `agmsg` scripts.\n2. Create the `.orchestration` directories before assigning work.\n3. Write a task file that includes objective, scope, allowed files, forbidden actions, expected artifacts, validation commands, and max turns.\n4. Start worker panes if needed. With herdr, send the worker prompt text into the pane and submit it with a carriage return; sending text alone may leave it unsubmitted.\n5. Configure delivery deliberately. `delivery.sh set turn` is useful for turn-end inbox checks; changing delivery mode can kill project watcher processes, so do it before starting long-running project watchers.\n6. Send `AGMSG-TASK v1` with the exact artifact paths and `done_signal=AGMSG-RESULT`.\n7. Track `max_turns`. Use `AGMSG-PING` for liveness if a worker stalls.\n8. On `AGMSG-RESULT`, read the task file and every referenced artifact before deciding.\n9. Send `AGMSG-ACCEPTANCE v1 status=accepted` when done, or `status=revise` with a narrow `reason` and `next_action` when more work is required.\n\n## Worker Playbook\n\n1. Read the full `AGMSG-TASK v1` message.\n2. Switch to the `repo` and read `task_file` before editing or running validations.\n3. Treat `allowed_files` as the edit boundary. If it says to see the task file, read that section and follow it exactly.\n4. Do not perform any `forbidden_actions`.\n5. Write artifacts to the exact expected paths. Do not invent alternate paths.\n6. Put command outputs and validation evidence in `expected_validation_file`.\n7. Put sandbox\u002FOpenSandbox status or fallback rationale in `expected_sandbox_file`.\n8. Put reusable learning triage in `expected_learning_file`; do not promote rules directly unless the task explicitly allows it.\n9. Put AutoSkill run status or a not-used record in `expected_autoskill_file`.\n10. If blocked, still write the report and evidence paths that explain the blocker.\n11. Reply with the requested `done_signal`, normally `AGMSG-RESULT v1`, and include all artifact paths.\n\n## Pitfalls\n\n- Do not start work from the agmsg message alone; read `task_file` first.\n- Do not edit outside `allowed_files`, even for convenient cleanup.\n- Do not perform forbidden actions such as dependency changes, gate changes, product changes, promotion decisions, image builds, or LLM calls when listed.\n- Do not collapse candidate, promoted, rejected, and merged skill registry states into one directory.\n- Do not put secrets, raw logs with credentials, or unredacted AutoSkill inputs in artifacts.\n- Do not install Hermes Agents runtime for this protocol.\n- Do not assume `herdr` text injection submits automatically; send Enter\u002FCR.\n- Do not treat `AGMSG-ACCEPTANCE status=revise` as a new task unless it changes the task file or explicitly provides a next action.\n",{"name":5,"metadata":23,"description":7},{"owner":9,"scope":24,"freshness":25,"provenance":26,"last-verified":27},"org","90d","repo=mryfmo\u002Fdotfiles; license=MIT",1784592000000,"main","8b4cc13bc73f8d52e6aa54d860c9d345b31d4032",".claude\u002Fskills\u002Fagmsg-orchestration","https:\u002F\u002Fgithub.com\u002Fmryfmo\u002Fherness-self-improvement","https:\u002F\u002Fgithub.com\u002Fmryfmo","2026-07-21 09:19:35",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add mryfmo\u002Fherness-self-improvement --skill agmsg-orchestration",{"tool":40,"command":41},"bunx","bunx skills add mryfmo\u002Fherness-self-improvement --skill agmsg-orchestration",{"tool":43,"command":44},"pnpm","pnpm dlx skills add mryfmo\u002Fherness-self-improvement --skill agmsg-orchestration","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fmryfmo\u002Fherness-self-improvement\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fagmsg-orchestration\u002FSKILL.md",true,[48],{"path":49,"resourceType":50,"resourceTypeDesc":51,"sizeBytes":52,"sourceHtmlUrl":53},"eval\u002FNOTES.md","REFERENCE","参考文档",168,"https:\u002F\u002Fgithub.com\u002Fmryfmo\u002Fherness-self-improvement\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fagmsg-orchestration\u002Feval\u002FNOTES.md",[55,62,69,76,83,90],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":18},"01M79BBVDZ6CHDG1Z9GMV5EJA6","shdoc-shell-docs","Write and review shellscript documentation with shdoc annotations. Use when Codex creates, edits, or reviews `.sh` files or shell executables and should add, repair, or normalize `@file`, `@brief`, `@description`, `@arg`, `@option`, and `@example` comments to match shdoc conventions.","https:\u002F\u002Fgithub.com\u002Fmryfmo\u002Fherness-self-improvement\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fshdoc-shell-docs\u002FSKILL.md",[],[],{"skillNo":63,"name":64,"nameZh":6,"description":65,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":18},"01M79BBRPXW9ZSYKP02N6K7M4G","python-uv-workflow","Apply Python development policy using uv-first execution, test-first behavior validation, and pre-commit quality gates. Use when implementing or refactoring Python code.","https:\u002F\u002Fgithub.com\u002Fmryfmo\u002Fherness-self-improvement\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fpython-uv-workflow\u002FSKILL.md",[],[],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":18},"01M79BBNZB6B9R8KNQV5V4SCJM","humanizer-ja","AIくさい日本語を、人が書いたように自然でこなれた日本語へ書き換えるスキル。ブログ記事、メモ、メール、文書、SNS投稿、チャット返信などの日本語を書いたり推敲したりするときに使う。AIっぽい言い回しを消したいとき、日本語を自然にしたいとき、言い換えたいとき、硬さを抜きたいとき、トーンを整えたいとき、人間らしい声や温度を足したいときに使う。","https:\u002F\u002Fgithub.com\u002Fmryfmo\u002Fherness-self-improvement\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fhumanizer-ja\u002FSKILL.md",[],[],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":18},"01M79BBJPWD4SH8EFC0CAWH64H","gh-first-workflow","Enforce gh-first GitHub investigation, pull request maintenance, and Conventional Commit output rules. Use when investigating GitHub issues or pull requests, creating or updating pull requests, summarizing investigation results, or preparing commit messages.","https:\u002F\u002Fgithub.com\u002Fmryfmo\u002Fherness-self-improvement\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fgh-first-workflow\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":18},"01M79BBF1KAC97FM9T2PDB4Z4B","gh-comment-attach-files","Attach local files to a GitHub issue or pull request comment via Playwright CLI and return the hosted attachment URLs without submitting the comment. Use when you need GitHub-hosted image or document URLs on github.com or GitHub Enterprise Server.","https:\u002F\u002Fgithub.com\u002Fmryfmo\u002Fherness-self-improvement\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fgh-comment-attach-files\u002FSKILL.md",[],[],{"skillNo":91,"name":92,"nameZh":6,"description":93,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":96,"publishedAt":18},"01M79BB8HF9FBW7HVEM080Q2NK","convert-to-transformers","Use this skill when converting custom PyTorch models to Hugging Face Transformers format. Helps with: (1) Creating PretrainedConfig and PreTrainedModel classes, (2) Writing ImageProcessor\u002FTokenizer, (3) Compatibility testing, (4) Hub upload preparation. Use when the user wants to make their model compatible with transformers library.","https:\u002F\u002Fgithub.com\u002Fmryfmo\u002Fherness-self-improvement\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fconvert-to-transformers\u002FSKILL.md",[],[]]