[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78DA7QZ2SA44J35N1ARVVPD":3,"skill-same:01M78DA7QZ2SA44J35N1ARVVPD":79},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":6,"ownerHtmlUrl":30,"ownerBlogUrl":6,"ownerDescription":31,"updatedAt":32,"install":33,"resources":46},"01M78DA7QZ2SA44J35N1ARVVPD","tailscale-wif",null,"Set up and troubleshoot SECRETLESS GitHub Actions -> Tailscale authentication using Workload Identity Federation (WIF): each run mints a GitHub OIDC token that a Tailscale \"Trust Credential\" exchanges for a short-lived API token or ephemeral auth key, so no Tailscale API key or OAuth client secret is ever stored. Use this WHENEVER a GitHub Actions workflow needs to talk to Tailscale — pushing an ACL\u002Fpolicy file with tailscale\u002Fgitops-acl-action, or joining a CI runner to the tailnet with tailscale\u002Fgithub-action — and you want to drop the stored credential, OR when such a workflow is failing with \"token exchange failed with status 403: Unauthorized\", \"requested tags are invalid or not permitted\", or \"only one of API Key, OAuth secret, or OAuth client ID and audience\". Reach for it on phrases like \"Tailscale OIDC\", \"Tailscale WIF\", \"federated identity Tailscale\", \"secretless Tailscale in CI\", \"remove the Tailscale API key \u002F OAuth secret from Actions\", or \"Tailscale Trust Credential\".","JRichlen\u002Fagent-plugins","JRichlen","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F9574264?v=4","01M78D3QWDTGRFSC2A9XADS6GM","Jordan Richlen","01M78D3QX3W2ZQEEACQHKDX3HW",1,"MIT","https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftailscale-wif\u002Fskills\u002Ftailscale-wif\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: tailscale-wif\ndescription: >-\n  Set up and troubleshoot SECRETLESS GitHub Actions -> Tailscale authentication\n  using Workload Identity Federation (WIF): each run mints a GitHub OIDC token\n  that a Tailscale \"Trust Credential\" exchanges for a short-lived API token or\n  ephemeral auth key, so no Tailscale API key or OAuth client secret is ever\n  stored. Use this WHENEVER a GitHub Actions workflow needs to talk to Tailscale\n  — pushing an ACL\u002Fpolicy file with tailscale\u002Fgitops-acl-action, or joining a CI\n  runner to the tailnet with tailscale\u002Fgithub-action — and you want to drop the\n  stored credential, OR when such a workflow is failing with \"token exchange\n  failed with status 403: Unauthorized\", \"requested tags are invalid or not\n  permitted\", or \"only one of API Key, OAuth secret, or OAuth client ID and\n  audience\". Reach for it on phrases like \"Tailscale OIDC\", \"Tailscale WIF\",\n  \"federated identity Tailscale\", \"secretless Tailscale in CI\", \"remove the\n  Tailscale API key \u002F OAuth secret from Actions\", or \"Tailscale Trust Credential\".\nlicense: MIT\ncompatibility: >-\n  Everything here is plain GitHub Actions YAML plus `gh`\u002F`curl`\u002F`jq` — it is\n  fully harness-agnostic. The bundled command entry point and the optional\n  Claude Code agent under agents\u002F are conveniences; on any other harness, read\n  this SKILL.md and run the same steps and scripts directly. PORTABILITY: the\n  setup\u002Ftroubleshoot procedure ports to any harness — only the invocation wrapper\n  is Claude-specific.\n---\n\n# Tailscale Workload Identity Federation (WIF) for GitHub Actions\n\n## Invariant\n\nWIF auth is **secretless**. Only the **public** `oauth-client-id` and `audience`\nare stored — and they belong in repo **variables**, never secrets. The job that\nauthenticates to Tailscale always carries `permissions: id-token: write`, and it\npasses **exactly one** auth mode. No long-lived Tailscale credential (a REST API\nkey, an OAuth client secret) is ever stored to enable a run: each run mints a\nfresh GitHub OIDC token, and a Tailscale Trust Credential exchanges it for a\nshort-lived (~1h) API token or an ephemeral auth key that expires on its own.\n\nIf a change would put a Tailscale secret back into the workflow, drop\n`id-token: write`, or pass two auth modes at once, it breaks this plugin's whole\nreason to exist. The cheap eval (`evals\u002Fcheap\u002Fchecks.sh`) defends the marker\nfacts below.\n\n## Why WIF\n\nStored Tailscale credentials are the thing you are removing:\n\n- **REST API keys** expire in **90 days or less** — someone has to rotate them,\n  and a leaked key is valid until it expires.\n- **OAuth client secrets** are long-lived bearer secrets — a leak is worse.\n\nWIF (a GA Tailscale feature) replaces both. Trust flows the other way: instead of\nhanding Tailscale a stored secret, each workflow run proves its identity with a\nGitHub-signed **OIDC token**, and Tailscale's **Trust Credential** (configured in\nthe admin console) decides whether to trust that identity and, if so, mints a\nshort-lived token. Because the `client-id` and `audience` are just **public\nidentifiers** (not secrets), they go in repo **variables** — nothing sensitive is\nstored anywhere.\n\n## The two use cases\n\nPick the one that matches what the workflow does with Tailscale.\n\n### 1. ACL \u002F policy GitOps — `tailscale\u002Fgitops-acl-action`\n\nPush your tailnet policy file from CI. Pin the action to a **SHA** (not a moving\ntag). The Trust Credential's **scope is `policy_file`**. The action fetches the\nGitHub OIDC token itself, so you only give it the client-id and audience.\n\nTemplate: `skills\u002Ftailscale-wif\u002Ftemplates\u002Facl-gitops.yml`\n\nKey change from the old secret-based setup: **remove `api-key`**, add\n`oauth-client-id` + `audience`.\n\n### 2. Node join (a CI runner joins the tailnet) — `tailscale\u002Fgithub-action`\n\nA runner brings up `tailscaled` and joins your tailnet for the duration of the\njob. Requires **`tailscale\u002Fgithub-action` v4.1.3+** for federated identity, and\n**`tailscaled` >= 1.90.1** for WIF — set `version: latest` so the installed\ndaemon is new enough. The Trust Credential's **scope is `auth_keys`**, and it\nmust be allowed to **assign the tag(s)** the runner needs (e.g. `tag:ci`).\n\nTemplate: `skills\u002Ftailscale-wif\u002Ftemplates\u002Fnode-join.yml`\n\nKey change from the old setup: **remove `oauth-secret`**, add `oauth-client-id` +\n`audience` + `tags` + `version: latest`. The tag you pass (e.g. `tag:ci`) must\n**also exist in the tailnet policy `tagOwners`**.\n\n## Workflow requirements (both use cases)\n\n1. **`permissions: id-token: write`** on the job (plus `contents: read` for\n   `actions\u002Fcheckout`). Without `id-token: write` the runner cannot mint the\n   GitHub OIDC token, so the exchange fails before it starts.\n2. **Exactly one auth mode.** `api-key` **XOR** `oauth-secret` **XOR**\n   (`oauth-client-id` **+** `audience`). Passing more than one makes the action\n   error: *\"only one of API Key, OAuth secret, or OAuth client ID and audience\"*.\n3. Store `oauth-client-id` and `audience` as **repo variables** (e.g.\n   `vars.TS_OIDC_CLIENT_ID`, `vars.TS_OIDC_AUDIENCE`). They are public — do not\n   waste secrets on them, and do not put a Tailscale secret next to them.\n\n## Trust Credential setup (Tailscale admin console)\n\nAdmin console -> **Trust credentials** -> **Credential** -> **OpenID Connect**.\nFull annotated walkthrough:\n`skills\u002Ftailscale-wif\u002Freferences\u002Ftrust-credential-console-setup.md`.\n\nFields:\n\n- **Issuer** — choose **\"GitHub Actions\"**; the console fills in\n  `token.actions.githubusercontent.com`.\n- **Subject** — the GitHub OIDC `sub` claim, with a trailing **`:*`** so one\n  credential matches **both** `pull_request` (the test\u002Fplan runs) **and**\n  `ref:refs\u002Fheads\u002Fmain` (the merge\u002Fapply run). See the subject-hardening section\n  below — for many orgs the correct value is the **ID-embedded** form.\n- **Custom claims** — leave empty unless you are deliberately narrowing further.\n- **Scope** — **`policy_file`** (ACL GitOps) or **`auth_keys`** (node join). For\n  `auth_keys`, also assign the **Tag** (e.g. `tag:ci`); that tag must **also**\n  appear in the tailnet policy `tagOwners`, or the join fails.\n- **Audience** — **generated by Tailscale** as `api.tailscale.com\u002F\u003Cclient-id>`.\n  You do **not** type your own audience.\n- **Output** — a **Client ID** and an **Audience**. Store both as repo variables\n  (`TS_OIDC_CLIENT_ID`, `TS_OIDC_AUDIENCE`).\n\n## The big gotcha — OIDC subject hardening\n\nDeep dive: `skills\u002Ftailscale-wif\u002Freferences\u002Foidc-subject-hardening.md`.\n\nMost examples show a **name-only** subject:\n\n```\nrepo:ORG\u002FREPO:pull_request\n```\n\nBut some orgs\u002Fenterprises **harden** their OIDC config to embed the numeric\norg\u002Frepo **database IDs** in the `sub`:\n\n```\nrepo:ORG@306084426\u002FREPO@1304364146:pull_request\n```\n\nIf your org does this and you configure a **name-only** subject match, the\nexchange **silently returns 403** — `token exchange failed with status 403:\nUnauthorized` — because the `sub` Tailscale sees never matches what you typed.\n\n**Fix:** match the **ID-embedded** form with a trailing `:*`:\n\n```\nrepo:ORG@\u003CorgId>\u002FREPO@\u003CrepoId>:*\n```\n\nThis is also **rename-proof** (numeric IDs don't change when you rename the org\nor repo), so it is the better match even when the name-only form would work. Get\nthe IDs with:\n\n```sh\ngh api repos\u002FOWNER\u002FREPO --jq '{id, owner_id: .owner.id}'\n```\n\n`.id` is the **repo** ID (`@\u003CrepoId>`); `.owner_id` is the **org\u002Fowner** ID\n(`@\u003CorgId>`). The helper `skills\u002Ftailscale-wif\u002Fscripts\u002Fcompute-subject.sh\nOWNER\u002FREPO` prints the exact ID-embedded subject to paste into the console.\n\n### Debug technique — decode the actual `sub`\n\nWhen in doubt, find out what GitHub is really signing. Add a **temporary**\nworkflow step (the job needs `id-token: write`) that mints a token and decodes\nits payload:\n\n```sh\n# In a workflow `run:` step (the job needs id-token: write). Set AUD from your\n# audience repo variable — the ready-made template wires it via the vars context.\nAUD=\"api.tailscale.com\u002F\u003Cyour-client-id>\"\nTOK=$(curl -sH \"Authorization: Bearer $ACTIONS_ID_TOKEN_REQUEST_TOKEN\" \\\n  \"$ACTIONS_ID_TOKEN_REQUEST_URL&audience=$AUD\" | jq -r .value)\nPAYLOAD=$(echo \"$TOK\" | cut -d. -f2)                 # JWT payload = 2nd dot-segment\ncase $(( ${#PAYLOAD} % 4 )) in 2) PAYLOAD=\"$PAYLOAD==\";; 3) PAYLOAD=\"$PAYLOAD=\";; esac  # pad to len%4\necho \"$PAYLOAD\" | tr '_-' '\u002F+' | base64 -d | jq '{iss, aud, sub}'   # base64url -> base64 -> decode\n```\n\nThe `sub` it prints is **exactly** what your Trust Credential's Subject must\nmatch. A ready-made standalone job (with the audience wired from the repo\nvariable) is `skills\u002Ftailscale-wif\u002Ftemplates\u002Fdebug-oidc-sub.yml` — delete the\nstep once the real subject is confirmed.\n\n## Symptom -> cause table\n\n| Symptom | Cause | Fix |\n|---|---|---|\n| `token exchange failed with status 403: Unauthorized` | Subject mismatch (usually OIDC subject hardening \u002F wrong IDs), or wrong audience | Decode the real `sub` (above); set the Trust Credential Subject to the **ID-embedded** `repo:ORG@\u003CorgId>\u002FREPO@\u003CrepoId>:*`; confirm audience is `api.tailscale.com\u002F\u003Cclient-id>` |\n| `requested tags [tag:X] are invalid or not permitted` | The tag is missing from the tailnet policy `tagOwners`, **or** the Trust Credential isn't scoped to assign that tag | Add `tag:X` to `tagOwners` in the policy file **and** assign it on the credential (scope `auth_keys`) |\n| `only one of API Key, OAuth secret, or OAuth client ID and audience` | More than one auth mode passed to the action | Pass **exactly one** mode; for WIF, remove `api-key` \u002F `oauth-secret` and keep only `oauth-client-id` + `audience` |\n\n## Workflow to follow — set up\n\n1. **Identify the use case** — ACL GitOps (`policy_file`) or node join\n   (`auth_keys`). This decides the scope and the template.\n2. **Compute the subject.** Run `skills\u002Ftailscale-wif\u002Fscripts\u002Fcompute-subject.sh OWNER\u002FREPO` (or the\n   `gh api` command above) to get the ID-embedded\n   `repo:ORG@\u003CorgId>\u002FREPO@\u003CrepoId>:*`.\n3. **Create the Trust Credential** in the admin console with Issuer = GitHub\n   Actions, the subject from step 2, the correct scope (and tag for `auth_keys`).\n   Copy out the generated **Client ID** and **Audience**.\n4. **Store the two public values** as repo variables `TS_OIDC_CLIENT_ID` and\n   `TS_OIDC_AUDIENCE` (`gh variable set TS_OIDC_CLIENT_ID --body \"...\"`).\n5. **Edit the workflow** from the matching template: add `permissions:\n   id-token: write`, switch to `oauth-client-id` + `audience`, and **remove** the\n   old `api-key` \u002F `oauth-secret`. For node join add `tags:` and `version: latest`.\n6. **Verify** on a PR and on `main`. If it 403s, use the decode-the-sub debug\n   step, then reconcile the Subject.\n\n## Workflow to follow — troubleshoot\n\n1. Read the exact error and map it with the **symptom -> cause table**.\n2. For any 403, **decode the real `sub`** with the debug step and compare it to\n   the Trust Credential Subject character-for-character. The usual culprit is\n   subject hardening — switch to the ID-embedded form.\n3. For a tags error, check **both** `tagOwners` in the policy **and** the\n   credential's assigned tag.\n4. For the \"only one of…\" error, grep the workflow for a leftover `api-key:` or\n   `oauth-secret:` and delete it.\n5. Confirm the invariant still holds: no stored Tailscale secret, `id-token:\n   write` present, exactly one auth mode.\n\nFull troubleshooting reference:\n`skills\u002Ftailscale-wif\u002Freferences\u002Ftroubleshooting.md`.\n","# Tailscale Workload Identity Federation (WIF) for GitHub Actions\n\n## Invariant\n\nWIF auth is **secretless**. Only the **public** `oauth-client-id` and `audience`\nare stored — and they belong in repo **variables**, never secrets. The job that\nauthenticates to Tailscale always carries `permissions: id-token: write`, and it\npasses **exactly one** auth mode. No long-lived Tailscale credential (a REST API\nkey, an OAuth client secret) is ever stored to enable a run: each run mints a\nfresh GitHub OIDC token, and a Tailscale Trust Credential exchanges it for a\nshort-lived (~1h) API token or an ephemeral auth key that expires on its own.\n\nIf a change would put a Tailscale secret back into the workflow, drop\n`id-token: write`, or pass two auth modes at once, it breaks this plugin's whole\nreason to exist. The cheap eval (`evals\u002Fcheap\u002Fchecks.sh`) defends the marker\nfacts below.\n\n## Why WIF\n\nStored Tailscale credentials are the thing you are removing:\n\n- **REST API keys** expire in **90 days or less** — someone has to rotate them,\n  and a leaked key is valid until it expires.\n- **OAuth client secrets** are long-lived bearer secrets — a leak is worse.\n\nWIF (a GA Tailscale feature) replaces both. Trust flows the other way: instead of\nhanding Tailscale a stored secret, each workflow run proves its identity with a\nGitHub-signed **OIDC token**, and Tailscale's **Trust Credential** (configured in\nthe admin console) decides whether to trust that identity and, if so, mints a\nshort-lived token. Because the `client-id` and `audience` are just **public\nidentifiers** (not secrets), they go in repo **variables** — nothing sensitive is\nstored anywhere.\n\n## The two use cases\n\nPick the one that matches what the workflow does with Tailscale.\n\n### 1. ACL \u002F policy GitOps — `tailscale\u002Fgitops-acl-action`\n\nPush your tailnet policy file from CI. Pin the action to a **SHA** (not a moving\ntag). The Trust Credential's **scope is `policy_file`**. The action fetches the\nGitHub OIDC token itself, so you only give it the client-id and audience.\n\nTemplate: `skills\u002Ftailscale-wif\u002Ftemplates\u002Facl-gitops.yml`\n\nKey change from the old secret-based setup: **remove `api-key`**, add\n`oauth-client-id` + `audience`.\n\n### 2. Node join (a CI runner joins the tailnet) — `tailscale\u002Fgithub-action`\n\nA runner brings up `tailscaled` and joins your tailnet for the duration of the\njob. Requires **`tailscale\u002Fgithub-action` v4.1.3+** for federated identity, and\n**`tailscaled` >= 1.90.1** for WIF — set `version: latest` so the installed\ndaemon is new enough. The Trust Credential's **scope is `auth_keys`**, and it\nmust be allowed to **assign the tag(s)** the runner needs (e.g. `tag:ci`).\n\nTemplate: `skills\u002Ftailscale-wif\u002Ftemplates\u002Fnode-join.yml`\n\nKey change from the old setup: **remove `oauth-secret`**, add `oauth-client-id` +\n`audience` + `tags` + `version: latest`. The tag you pass (e.g. `tag:ci`) must\n**also exist in the tailnet policy `tagOwners`**.\n\n## Workflow requirements (both use cases)\n\n1. **`permissions: id-token: write`** on the job (plus `contents: read` for\n   `actions\u002Fcheckout`). Without `id-token: write` the runner cannot mint the\n   GitHub OIDC token, so the exchange fails before it starts.\n2. **Exactly one auth mode.** `api-key` **XOR** `oauth-secret` **XOR**\n   (`oauth-client-id` **+** `audience`). Passing more than one makes the action\n   error: *\"only one of API Key, OAuth secret, or OAuth client ID and audience\"*.\n3. Store `oauth-client-id` and `audience` as **repo variables** (e.g.\n   `vars.TS_OIDC_CLIENT_ID`, `vars.TS_OIDC_AUDIENCE`). They are public — do not\n   waste secrets on them, and do not put a Tailscale secret next to them.\n\n## Trust Credential setup (Tailscale admin console)\n\nAdmin console -> **Trust credentials** -> **Credential** -> **OpenID Connect**.\nFull annotated walkthrough:\n`skills\u002Ftailscale-wif\u002Freferences\u002Ftrust-credential-console-setup.md`.\n\nFields:\n\n- **Issuer** — choose **\"GitHub Actions\"**; the console fills in\n  `token.actions.githubusercontent.com`.\n- **Subject** — the GitHub OIDC `sub` claim, with a trailing **`:*`** so one\n  credential matches **both** `pull_request` (the test\u002Fplan runs) **and**\n  `ref:refs\u002Fheads\u002Fmain` (the merge\u002Fapply run). See the subject-hardening section\n  below — for many orgs the correct value is the **ID-embedded** form.\n- **Custom claims** — leave empty unless you are deliberately narrowing further.\n- **Scope** — **`policy_file`** (ACL GitOps) or **`auth_keys`** (node join). For\n  `auth_keys`, also assign the **Tag** (e.g. `tag:ci`); that tag must **also**\n  appear in the tailnet policy `tagOwners`, or the join fails.\n- **Audience** — **generated by Tailscale** as `api.tailscale.com\u002F\u003Cclient-id>`.\n  You do **not** type your own audience.\n- **Output** — a **Client ID** and an **Audience**. Store both as repo variables\n  (`TS_OIDC_CLIENT_ID`, `TS_OIDC_AUDIENCE`).\n\n## The big gotcha — OIDC subject hardening\n\nDeep dive: `skills\u002Ftailscale-wif\u002Freferences\u002Foidc-subject-hardening.md`.\n\nMost examples show a **name-only** subject:\n\n```\nrepo:ORG\u002FREPO:pull_request\n```\n\nBut some orgs\u002Fenterprises **harden** their OIDC config to embed the numeric\norg\u002Frepo **database IDs** in the `sub`:\n\n```\nrepo:ORG@306084426\u002FREPO@1304364146:pull_request\n```\n\nIf your org does this and you configure a **name-only** subject match, the\nexchange **silently returns 403** — `token exchange failed with status 403:\nUnauthorized` — because the `sub` Tailscale sees never matches what you typed.\n\n**Fix:** match the **ID-embedded** form with a trailing `:*`:\n\n```\nrepo:ORG@\u003CorgId>\u002FREPO@\u003CrepoId>:*\n```\n\nThis is also **rename-proof** (numeric IDs don't change when you rename the org\nor repo), so it is the better match even when the name-only form would work. Get\nthe IDs with:\n\n```sh\ngh api repos\u002FOWNER\u002FREPO --jq '{id, owner_id: .owner.id}'\n```\n\n`.id` is the **repo** ID (`@\u003CrepoId>`); `.owner_id` is the **org\u002Fowner** ID\n(`@\u003CorgId>`). The helper `skills\u002Ftailscale-wif\u002Fscripts\u002Fcompute-subject.sh\nOWNER\u002FREPO` prints the exact ID-embedded subject to paste into the console.\n\n### Debug technique — decode the actual `sub`\n\nWhen in doubt, find out what GitHub is really signing. Add a **temporary**\nworkflow step (the job needs `id-token: write`) that mints a token and decodes\nits payload:\n\n```sh\n# In a workflow `run:` step (the job needs id-token: write). Set AUD from your\n# audience repo variable — the ready-made template wires it via the vars context.\nAUD=\"api.tailscale.com\u002F\u003Cyour-client-id>\"\nTOK=$(curl -sH \"Authorization: Bearer $ACTIONS_ID_TOKEN_REQUEST_TOKEN\" \\\n  \"$ACTIONS_ID_TOKEN_REQUEST_URL&audience=$AUD\" | jq -r .value)\nPAYLOAD=$(echo \"$TOK\" | cut -d. -f2)                 # JWT payload = 2nd dot-segment\ncase $(( ${#PAYLOAD} % 4 )) in 2) PAYLOAD=\"$PAYLOAD==\";; 3) PAYLOAD=\"$PAYLOAD=\";; esac  # pad to len%4\necho \"$PAYLOAD\" | tr '_-' '\u002F+' | base64 -d | jq '{iss, aud, sub}'   # base64url -> base64 -> decode\n```\n\nThe `sub` it prints is **exactly** what your Trust Credential's Subject must\nmatch. A ready-made standalone job (with the audience wired from the repo\nvariable) is `skills\u002Ftailscale-wif\u002Ftemplates\u002Fdebug-oidc-sub.yml` — delete the\nstep once the real subject is confirmed.\n\n## Symptom -> cause table\n\n| Symptom | Cause | Fix |\n|---|---|---|\n| `token exchange failed with status 403: Unauthorized` | Subject mismatch (usually OIDC subject hardening \u002F wrong IDs), or wrong audience | Decode the real `sub` (above); set the Trust Credential Subject to the **ID-embedded** `repo:ORG@\u003CorgId>\u002FREPO@\u003CrepoId>:*`; confirm audience is `api.tailscale.com\u002F\u003Cclient-id>` |\n| `requested tags [tag:X] are invalid or not permitted` | The tag is missing from the tailnet policy `tagOwners`, **or** the Trust Credential isn't scoped to assign that tag | Add `tag:X` to `tagOwners` in the policy file **and** assign it on the credential (scope `auth_keys`) |\n| `only one of API Key, OAuth secret, or OAuth client ID and audience` | More than one auth mode passed to the action | Pass **exactly one** mode; for WIF, remove `api-key` \u002F `oauth-secret` and keep only `oauth-client-id` + `audience` |\n\n## Workflow to follow — set up\n\n1. **Identify the use case** — ACL GitOps (`policy_file`) or node join\n   (`auth_keys`). This decides the scope and the template.\n2. **Compute the subject.** Run `skills\u002Ftailscale-wif\u002Fscripts\u002Fcompute-subject.sh OWNER\u002FREPO` (or the\n   `gh api` command above) to get the ID-embedded\n   `repo:ORG@\u003CorgId>\u002FREPO@\u003CrepoId>:*`.\n3. **Create the Trust Credential** in the admin console with Issuer = GitHub\n   Actions, the subject from step 2, the correct scope (and tag for `auth_keys`).\n   Copy out the generated **Client ID** and **Audience**.\n4. **Store the two public values** as repo variables `TS_OIDC_CLIENT_ID` and\n   `TS_OIDC_AUDIENCE` (`gh variable set TS_OIDC_CLIENT_ID --body \"...\"`).\n5. **Edit the workflow** from the matching template: add `permissions:\n   id-token: write`, switch to `oauth-client-id` + `audience`, and **remove** the\n   old `api-key` \u002F `oauth-secret`. For node join add `tags:` and `version: latest`.\n6. **Verify** on a PR and on `main`. If it 403s, use the decode-the-sub debug\n   step, then reconcile the Subject.\n\n## Workflow to follow — troubleshoot\n\n1. Read the exact error and map it with the **symptom -> cause table**.\n2. For any 403, **decode the real `sub`** with the debug step and compare it to\n   the Trust Credential Subject character-for-character. The usual culprit is\n   subject hardening — switch to the ID-embedded form.\n3. For a tags error, check **both** `tagOwners` in the policy **and** the\n   credential's assigned tag.\n4. For the \"only one of…\" error, grep the workflow for a leftover `api-key:` or\n   `oauth-secret:` and delete it.\n5. Confirm the invariant still holds: no stored Tailscale secret, `id-token:\n   write` present, exactly one auth mode.\n\nFull troubleshooting reference:\n`skills\u002Ftailscale-wif\u002Freferences\u002Ftroubleshooting.md`.\n",{"name":5,"license":15,"description":7,"compatibility":24},"Everything here is plain GitHub Actions YAML plus `gh`\u002F`curl`\u002F`jq` — it is fully harness-agnostic. The bundled command entry point and the optional Claude Code agent under agents\u002F are conveniences; on any other harness, read this SKILL.md and run the same steps and scripts directly. PORTABILITY: the setup\u002Ftroubleshoot procedure ports to any harness — only the invocation wrapper is Claude-specific.","main","929887017aa19a575de99be4e3f9b78d0d7ad48d","plugins\u002Ftailscale-wif\u002Fskills\u002Ftailscale-wif",0,"https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins","https:\u002F\u002Fgithub.com\u002FJRichlen","Solutions Architect","2026-11-12 14:00:54",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add JRichlen\u002Fagent-plugins --skill tailscale-wif",{"tool":39,"command":40},"bunx","bunx skills add JRichlen\u002Fagent-plugins --skill tailscale-wif",{"tool":42,"command":43},"pnpm","pnpm dlx skills add JRichlen\u002Fagent-plugins --skill tailscale-wif","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftailscale-wif\u002Fskills\u002Ftailscale-wif\u002FSKILL.md",true,[47,53,57,61,67,71,75],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Foidc-subject-hardening.md","REFERENCE","参考文档",2290,"https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftailscale-wif\u002Fskills\u002Ftailscale-wif\u002Freferences\u002Foidc-subject-hardening.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Ftroubleshooting.md",2139,"https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftailscale-wif\u002Fskills\u002Ftailscale-wif\u002Freferences\u002Ftroubleshooting.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Ftrust-credential-console-setup.md",3110,"https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftailscale-wif\u002Fskills\u002Ftailscale-wif\u002Freferences\u002Ftrust-credential-console-setup.md",{"path":62,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":65,"sourceHtmlUrl":66},"scripts\u002Fcompute-subject.sh","SCRIPT","脚本",2294,"https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftailscale-wif\u002Fskills\u002Ftailscale-wif\u002Fscripts\u002Fcompute-subject.sh",{"path":68,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":69,"sourceHtmlUrl":70},"templates\u002Facl-gitops.yml",1752,"https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftailscale-wif\u002Fskills\u002Ftailscale-wif\u002Ftemplates\u002Facl-gitops.yml",{"path":72,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":73,"sourceHtmlUrl":74},"templates\u002Fdebug-oidc-sub.yml",1518,"https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftailscale-wif\u002Fskills\u002Ftailscale-wif\u002Ftemplates\u002Fdebug-oidc-sub.yml",{"path":76,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":77,"sourceHtmlUrl":78},"templates\u002Fnode-join.yml",1612,"https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Ftailscale-wif\u002Fskills\u002Ftailscale-wif\u002Ftemplates\u002Fnode-join.yml",[80,87,94,101,108,115],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":19},"01M78DCJ7N0S5GVNPR0849HPRA","wayfinder","Chart a multi-session effort as a labeled map of typed decision tickets (grilling \u002F prototype \u002F research \u002F task) with explicit dependencies and an open frontier agents self-assign into. Plans; never executes. Use this skill whenever the user is working with wayfinder, and on phrases like \"chart this effort\", \"break this into tickets\", \"what's the frontier\", \"plan across sessions\", or \"map out the dependencies for this work\".","https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fwayfinder\u002Fskills\u002Fwayfinder\u002FSKILL.md",[],[],{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":93,"publishedAt":19},"01M78DCB8APFG6B3G7D34WZ2H9","wave-research","Staged, adversarial research that tries to break a thesis before believing it: waves, checkpoints, falsifiable sub-claims, rival hypotheses, an ACH matrix, source-incentive tags and a verified final thesis. Also carries Workflow templates for a one-pass research-and-verify fan-out. Use to stress-test or vet a contested claim; not for quick lookups.","https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fwave-research\u002Fskills\u002Fwave-research\u002FSKILL.md",[],[],{"skillNo":95,"name":96,"nameZh":6,"description":97,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":98,"categories":99,"keywords":100,"publishedAt":19},"01M78DC30012WTD0B2T21ZFVFG","second-opinion","Offer-only validation pipeline for a verdict or recommendation. Use when the user asks to validate, verify, fact-check, or get a second opinion on a claim or recommendation — \"are you sure\", \"double-check this\", \"second opinion\". Batches fact-checking against sources, adds scoped advisor personas for the angles that need judgment, then re-emits the response grouped into verified, flagged, and conflict, with a mandatory validation delta. Requires a subagent-spawning tool; without one it must decline rather than simulate.","https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fvoice\u002Fskills\u002Fsecond-opinion\u002FSKILL.md",[],[],{"skillNo":102,"name":103,"nameZh":6,"description":104,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":105,"categories":106,"keywords":107,"publishedAt":19},"01M78DBTTZAK0Z8FZ7RCFX92NM","machine-voice","Compression rules for standalone machine-read artifacts the reader scans — agent traces, tool-call logs, status lines, progress updates, handoff notes, state dumps, config, schemas, reference cards, and structured data another agent parses. Applies three layers: lexical compression, markdown structure as navigation, and emojis as typed status markers. Reach for it when emitting one of those artifacts on its own. Companion to human-voice, which governs prose a human reads and also owns any table or block that serves surrounding explanation. Neither skill applies to code or file contents the user will save, commit messages, creative writing, or turns that are only a clarifying question — those ship unstyled and uncompressed.","https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fvoice\u002Fskills\u002Fmachine-voice\u002FSKILL.md",[],[],{"skillNo":109,"name":110,"nameZh":6,"description":111,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":112,"categories":113,"keywords":114,"publishedAt":19},"01M78DBH7MPRD7ZNKVRA6JK76S","human-voice","Style for prose a human reads — answers, recommendations, explanations, research summaries, comparisons. Use on every conversational reply that contains prose for a person to read: triage intent, lead with the verdict, keep it scannable, tag claim confidence. Reach for it whenever answering a technical question, comparing options, or making a recommendation, even when the user says nothing about style. Companion to machine-voice, which instead governs standalone machine-read artifacts — agent traces, tool-call logs, status lines, progress updates, handoff notes, state dumps, config, schemas, reference cards, and structured data another agent parses. Neither skill applies to code or file contents, commit messages, creative writing, or turns that are only a clarifying question — those ship unstyled.","https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fvoice\u002Fskills\u002Fhuman-voice\u002FSKILL.md",[],[],{"skillNo":116,"name":117,"nameZh":6,"description":118,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":119,"categories":120,"keywords":121,"publishedAt":19},"01M78DB9587AD3Y46FYVE2V3T4","ai-writing-mistakes","Catch and repair the tells that mark prose as machine-written — the negation-flip (\"it's not X, it's Y\"), rule-of-three padding, hedge stacking, unearned intensifiers, filler verbs like delve and leverage, participle trailers (\"...ensuring seamless integration\"), sycophantic openers, empty closing summaries, and uniform em-dash\u002Fsentence rhythm. Use when drafting or revising any prose a human will read, including documents you were asked to author — READMEs, docs, PR bodies, release notes, emails — and when the user asks whether something \"sounds like AI\" or wants writing tightened. This is a wording pass, not a voice: human-voice still decides layout and confidence tags, machine-voice still owns traces and logs, and neither this skill nor those two touch code, commit messages, or quoted text.","https:\u002F\u002Fgithub.com\u002FJRichlen\u002Fagent-plugins\u002Fblob\u002Fmain\u002Fplugins\u002Fvoice\u002Fskills\u002Fai-writing-mistakes\u002FSKILL.md",[],[]]