MCPInstall

service-itsm-agentic-setup-itsm-agentforce-permset-assign

Resolve missing ITSM Intelligence invocable actions so a Fulfiller NGA agent can activate. Reads which of the three Core Fulfiller persona permsets (IncidentFulfiller, ProblemFulfillerPermSet, ChangeRequestFulfillerPermSet) are provisioned, then assigns the running user the selected persona (plus backing PSL when license-gated) so svcitsmintelligence actions surface; hands off to service-itsm-agentic-setup-agentforce-studio-validate if none are provisioned. Use when the Fulfiller agent-configure skill reports missing actions on activate, when 'Invocable action svcitsmintelligenceX does not exist' surfaces, when a user asks to grant themselves Fulfiller prompt-template access, when asked to assign the Incident, Problem, or Change Fulfiller persona permission set, or when resolving missing ITSM Intelligence action access for the Fulfiller agent. DO NOT TRIGGER for Employee-agent access, Agentforce for IT Service toggles, agent creation, CMDB access, or generic permset assignment

Installs
1.2K
Last 24h
—
Trending rank
—
Security
Not audited

Install

Works with every agent the skills CLI supports.

npx skills add https://github.com/forcedotcom/sf-skills --skill service-itsm-agentic-setup-itsm-agentforce-permset-assign
  1. 01

    Install

    Run the command in your project. The CLI asks which agents to install for.

  2. 02

    Describe the task

    Ask your agent for something this skill covers. It loads service-itsm-agentic-setup-itsm-agentforce-permset-assign automatically when relevant.

  3. 03

    Review the result

    The agent follows the SKILL.md steps, scripts and references bundled with the skill.

SKILL.md

The instructions your agent reads when the skill is triggered.

Assign an ITSM Fulfiller Persona Permission Set (Prompt-Template Access)

Grants the running user one of the Core-shipped Fulfiller persona permission sets that expose the svc_itsm_intelligence__* prompt-template invocable actions on the target org — the actions the Fulfiller NGA agent scripts reference via source: / target: generatePromptResponse://.... When those invocables are not surfaced by /services/data/v67.0/actions/custom/generatePromptResponse for the running user, the Fulfiller agent-configure skill's Phase 6 activate call returns HTTP 200 with a silent {success:false, messages:[{... "does not exist"}]} body and the agent never becomes usable. This skill fixes that gap by assigning the correct Fulfiller persona permset (and its backing license when one exists) — or, when no Fulfiller persona permset is provisioned on the org at all, hands off to the Agentforce Studio configure/validate skill so the ITSM AddOn(s) can be enabled first.

The three Fulfiller persona permsets, their AddOns, PSLs, and the userPerms they grant are documented in references/permset-topology.md. All are Core-shipped in namespace force — there is no managed-package namespaced permset for this feature.

Employee agent is out of scope. The Employee NGA agent's access model is separate (org-preferences + a different persona layer) and does not map onto these three persona permsets.

Every call runs through the Salesforce CLI (sf):

  • sf api request rest — authenticated Connect API GET (identity, verify read).
  • sf data query — SOQL on PermissionSet (persona presence), PermissionSetAssignment / PermissionSetLicenseAssign (idempotency).
  • sf org assign permset — assigning the permission set for the running user.
  • No token is ever extracted; no MCP is used.

Scope

  • In scope: detecting which of the three Fulfiller persona permsets (IncidentFulfiller, ProblemFulfillerPermSet, ChangeRequestFulfillerPermSet) are provisioned on the org, letting the user pick which persona to assign, checking existing assignments, assigning the permission-set license (when the persona is license-gated) and permission set to the running user (or a named user), verifying the target svc_itsm_intelligence__* invocable actions surface via a follow-up /actions/custom/generatePromptResponse read.
  • Out of scope: Employee-agent access (different access model, different skill), installing/enabling the ITSM AddOn(s) or content bundle (hand off to service-itsm-agentic-setup-agentforce-studio-validate), enabling org-level Agentforce feature toggles, creating a permission set, creating or activating the Fulfiller agent (that's service-itsm-agentic-setup-fulfiller-agent-configure), CMDB access (service-itsm-agentic-setup-cmdb-access-assign), generic non-ITSM permission-set assignment (dx-org-permission-set-assign).

Mechanism

Two branches, decided by a read-only detection step first:

  • Branch A — one or more Fulfiller persona permsets exist on the org. Ask the user which persona to assign (do not auto-select — a Fulfiller commonly needs only one). Idempotent assign: PSL first when the persona is license-gated, then permission set, verified by read-back and by a follow-up /actions/custom/generatePromptResponse read.
  • Branch B — none of the three Fulfiller persona permsets exist on the org. The ITSM AddOn(s) are not provisioned; permset-assign is a no-op. STOP and hand off to service-itsm-agentic-setup-agentforce-studio-validate so the AddOn(s) can be enabled first.

The two-branch shape is deliberate — the failure signature ("svc_itsm_intelligence__X does not exist" on activate) looks identical whether a persona is present-and-unassigned or the AddOn is absent entirely, and there is no way to tell from the activate response alone. The pre-check on the three persona PermissionSet names is what disambiguates them.

Four helper scripts (all invoked via Bash) hold every deterministic decision (A9). Full I/O contracts in references/helper-contracts.md; workflow-level usage summarized below:

  • scripts/classify-permset-availability.mjs — Branch A vs B and the per-persona needsPsl flag. Returns the full candidates[] (personas actually on the org) for the caller to prompt on.
  • scripts/resolve-target-user.mjs — extracts 005… running-user Id from the API-root identity URL. Fails closed on any malformed shape.
  • scripts/classify-assignment-state.mjs — idempotency; pass the sentinel NO-PSL in place of the PSLA path when the selected persona's needsPsl:false.
  • scripts/classify-action-surface.mjs — Phase 4 verify verdict from the /actions/custom/generatePromptResponse capture (with optional expected-actions CSV).

Preconditions

  1. sf CLI installed and authenticated to the target org (sf org display -o <alias> shows Connected). All calls use --target-org <alias>; never extract or pass the access token by hand.
  2. API v67.0+.
  3. node ≥ 18 on PATH.

If a precondition fails, sf surfaces an auth or 401/403/404; report the raw response verbatim and stop.


Clarifying questions

Ask only what cannot be inferred from conversation:

  • Target org — the sf alias. Default to sf config get target-org if unset.
  • Target user — default to the running user (resolved via scripts/resolve-target-user.mjs). If the user asks to assign on behalf of a named user, resolve them by Username first.
  • Which Fulfiller persona? Incident / Problem / Change. Only ask about personas that are actually provisioned on the org (from candidates[]). Do not auto-select — a Fulfiller commonly needs only one persona (e.g. Incident) even when others are provisioned.
  • Confirm the write — assigning a permission-set license consumes a seat and takes effect for a live user session. Present the target user + org + persona permset name, and require an explicit "yes" via AskUserQuestion before writing.

Workflow

All calls go through sf; substitute <alias> with the target org.

Phase 1 — Read: which Fulfiller persona permsets are provisioned on this org?

  1. Query PermissionSet for the three known Fulfiller persona DeveloperNames:

    sf data query \
      -q "SELECT Id, Name, Label, LicenseId FROM PermissionSet WHERE Name IN ('IncidentFulfiller','ProblemFulfillerPermSet','ChangeRequestFulfillerPermSet')" \
      --target-org <alias> --json > /tmp/itsm-personas.json 2>/tmp/itsm-personas.err || true
    

    (The PermissionSet namespace on all three is force — do NOT filter by NamespacePrefix.)

  2. Classify:

    node "<skill_dir>/scripts/classify-permset-availability.mjs" /tmp/itsm-personas.json
    

    The classifier prints { personasFound, personasMissing, candidates, verdict, reasons }, where each candidates[] row is {Id, Name, Label, LicenseId, needsPsl}:

    • verdict:"ASSIGN" (≥1 persona present) ⇒ continue to Phase 2. Present the personasFound list to the user via AskUserQuestion and get the selected persona; record its Id, LicenseId, and needsPsl — they drive whether Phase 2b/2d touch the PSL at all.
    • verdict:"HAND-OFF" (none of the three personas present) ⇒ Phase 2 is impossible on this org; go to Phase 3 (Branch B hand-off).
    • verdict:"CANNOT-CONFIRM" (query failed) ⇒ surface the raw CLI error verbatim; stop.

Phase 2 — Assign path (Branch A)

2a. Resolve the target user. Read the identity URL, then extract the user Id via the resolver (do NOT parse the URL by hand; do NOT use USER_ID() — Apex-only, rejected by REST; do NOT rely on /chatter/users/me — 403 when Chatter is off):

sf api request rest "/services/data/v67.0/" --method GET --target-org <alias> > /tmp/api-root.json 2>/tmp/api-root.err || true
node "<skill_dir>/scripts/resolve-target-user.mjs" /tmp/api-root.json

The resolver prints { userId, identity, verdict, reasons }. On verdict:"RESOLVED" use userId as the running user; on verdict:"CANNOT-CONFIRM" surface the reasons verbatim and stop — do NOT guess.

If the user asks to assign on behalf of a named user instead, resolve by Username:

sf data query \
  -q "SELECT Id, Username, Name, IsActive FROM User WHERE Username = '<username>'" \
  --target-org <alias> --json > /tmp/user-lookup.json 2>/tmp/user-lookup.err || true

2b. Idempotency read. Query the PermissionSetAssignment for the target user + selected persona's Id (SOQL shape in references/cli-invocation.md), and then branch on the selected persona's needsPsl:

  • needsPsl:true — query the PermissionSetLicenseAssign for the target user + the persona's LicenseId, then classify:

    node "<skill_dir>/scripts/classify-assignment-state.mjs" /tmp/psa-existing.json /tmp/psla-existing.json
    
  • needsPsl:false — skip the PSLA query entirely; pass the sentinel:

    node "<skill_dir>/scripts/classify-assignment-state.mjs" /tmp/psa-existing.json NO-PSL
    

The classifier prints { permsetAssigned, licenseAssigned, needsWrite, verdict, reasons }. If needsWrite:false ⇒ Phase 4 (verify only). Else continue to Phase 2c.

2c. Confirm-to-write checkpoint (REQUIRED). Present the target user + org + persona permset name and require an explicit "yes" via AskUserQuestion. On "no", stop and report the current state without any writes.

2d. Assign — order depends on the selected persona's needsPsl:

  • needsPsl:true — POST the PSL to /sobjects/PermissionSetLicenseAssign FIRST, then run sf org assign permset --name <permsetName> --on-behalf-of <userId>. Assigning the permission set without the PSL sticks the assignment but the license backing it never activates. Exact call shapes: references/cli-invocation.md.
  • needsPsl:false — SKIP the PSL POST entirely; run sf org assign permset only. (A persona whose PermissionSet has no backing LicenseId is not license-gated — there is no PSL seat to hold, so assigning the permset alone is the correct and complete write.) Retained defensively: every shipped persona is now PSL-backed (needsPsl is derived per-row from LicenseId), so this false branch and its NO-PSL wiring are currently unexercised by shipped data — kept for correctness against a future persona whose PermissionSet carries no backing LicenseId.

Response handling:

  • 201 on POST / success:true on sf org assign permset ⇒ assigned.
  • 400 DUPLICATE_VALUE on the PSL POST ⇒ user already had it; treat as success, not error.
  • 400 INSUFFICIENT_ACCESS / seat-exhaustion on the PSL POST ⇒ STOP for this write; tell the user the PSL has no seats available.

Phase 3 — Hand-off path (Branch B: no Fulfiller persona provisioned)

  1. When Phase 1 returns verdict:"HAND-OFF", none of the three Fulfiller persona permsets exist on this org — the ITSM AddOn(s) are not provisioned. Permset-assign is a no-op in this state. Present the discovery via AskUserQuestion:

    "None of the Fulfiller persona permission sets (Incident, Problem, Change) is provisioned on this org — no permset can grant access to actions that don't exist yet. Run service-itsm-agentic-setup-agentforce-studio-validate to diagnose which AddOn needs enabling?" (options: Yes, run the readiness check / No, stop here).

    • On Yes: delegate to service-itsm-agentic-setup-agentforce-studio-validate and let it recommend the configure/bundle-deploy skill.
    • On No: stop and report the current state (no persona provisioned, cannot assign) — no writes.

Phase 4 — Verify

  1. Regardless of write vs skip, re-read /actions/custom/generatePromptResponse and classify via the helper (never by prose grep — A9). CSV shape and both invocation forms live in references/helper-contracts.md / references/cli-invocation.md:

    sf api request rest "/services/data/v67.0/actions/custom/generatePromptResponse" \
      --method GET --target-org <alias> > /tmp/generate-prompt-response.json 2>/tmp/generate-prompt-response.err || true
    node "<skill_dir>/scripts/classify-action-surface.mjs" /tmp/generate-prompt-response.json [expectedActions-csv]
    

    The helper prints { present, missing, totalItsmActionsSeen, verdict, reasons }. On SURFACED proceed to Phase 5; on PARTIAL / MISSING after a successful assign, tell the user the write succeeded but the actions are not surfaced — session refresh or wrong persona. Do not falsely report success. On CANNOT-CONFIRM surface the reasons verbatim.

Phase 5 — Aggregate verdict

  1. Report one of:
    • ASSIGNED — Branch A wrote, verify saw the target actions surface.
    • ALREADY-ASSIGNED — Branch A found needsWrite:false; verify saw the target actions surface. (Assignment is already in place; no writes needed.)
    • HAND-OFF — Branch B; no Fulfiller persona is provisioned. Named the follow-up skill.
    • VERIFY-INCONCLUSIVE — write completed but the verify read didn't surface the expected action set. Surface the observed state verbatim; do not report success.
    • FAILED — any Phase 2d write returned an error other than DUPLICATE_VALUE. Report the raw error.

Rules / Constraints

ConstraintRationale
Detect Fulfiller persona presence via the three fixed PermissionSet.Name values BEFORE any permset writeThe failure signature ("action does not exist" on activate) is identical for AddOn-absent and permset-not-assigned; only the pre-check disambiguates them. The three personas are Core-shipped in namespace force — a NamespacePrefix filter never returns them
Ask the user which persona to assign — never auto-selectA Fulfiller commonly needs only one persona (e.g. Incident). Auto-assigning the first row returned would over-grant
All decisions are made by helper scripts, never by proseAssignment/idempotency logic is deterministic; prose interpretation is not (A9)
Assign the PSL before the permission set — ONLY when the selected persona's needsPsl:trueThe permission set is license-backed; the license seat must be held before the assignment sticks. When the selected persona has no LicenseId, needsPsl:false and the PSL POST is skipped entirely
Read needsPsl from the SELECTED persona's LicenseId, per-row — never from a namespace-wide PSL queryDifferent personas can have different license shapes on the same org; falling back to a namespace-wide PSL would POST a wrong PermissionSetLicenseAssign
Resolve the running-user Id via scripts/resolve-target-user.mjs — never by prose parsing the identity URLThe identity URL's segment shape (005…, 15 or 18 chars) is a hard rule; the classifier validates and fails closed
Classify the Phase 4 action surface via scripts/classify-action-surface.mjs — never by prose grep of the responseVerify is the gate for reporting SUCCESS vs VERIFY-INCONCLUSIVE; the decision must be deterministic
Treat 400 DUPLICATE_VALUE on PSL POST as successIt means the user already has that assignment — idempotent, not an error
Never create or edit a permission setThis skill only assigns the standard Fulfiller persona permission set(s); authoring perm sets is out of scope
Never install the AddOn / never toggle org-level Agentforce featuresThat is the service-itsm-agentic-setup-agentforce-studio-validate / -configure scope; this skill hands off, it does not enable
Verify after write via /actions/custom/generatePromptResponse — never trust POST return code aloneThe assignment can succeed while the target action surface still doesn't include what the Fulfiller template needs (wrong persona, cache)
Confirm-to-write checkpoint before Phase 2dA permset assign consumes a license seat and takes effect for a live user session
Never extract the access tokenUse sf api request rest / sf data query / sf org assign permset — they use the CLI's stored session
Report exact error text from the CLI responseEnables support to diagnose failures

Verification Checklist

  • Persona availability classified by scripts/classify-permset-availability.mjs against the three fixed persona Names — never by prose scanning the query output.
  • User was asked to pick a persona from personasFound[] — no auto-selection.
  • Target user Id resolved by scripts/resolve-target-user.mjs — never by prose splitting the identity URL.
  • The SELECTED persona's needsPsl drove Phase 2b/2d: PSL SOQL + POST were performed when true and skipped when false.
  • permsetLicenseId used for the PSL POST came from the SELECTED persona's own LicenseId — never a fallback from a namespace-wide query.
  • On Branch A: existing assignments read via sf data query before any write, classified by scripts/classify-assignment-state.mjs (with NO-PSL sentinel when needsPsl:false).
  • On Branch A: user confirmed the write at the Phase 2c checkpoint.
  • On Branch A + needsPsl:true: PSL was POSTed before the permission set was assigned.
  • DUPLICATE_VALUE on the PSL POST was treated as success, not failure.
  • On Branch B: no write was attempted; the hand-off to service-itsm-agentic-setup-agentforce-studio-validate was offered.
  • Phase 4 verify classified via scripts/classify-action-surface.mjs — no false ASSIGNED without the helper returning verdict:"SURFACED".
  • Aggregate verdict reported (ASSIGNED / ALREADY-ASSIGNED / HAND-OFF / VERIFY-INCONCLUSIVE / FAILED).

Output Format

ITSM Fulfiller Persona Permset Assignment (via service-itsm-agentic-setup-itsm-agentforce-permset-assign)

Org:            <org-alias> (API v67.0)
Target user:    <username> (<userId>)
Persona:        <Incident | Problem | Change>
PermSet:        <DeveloperName>

  Personas provisioned on org ...... <comma-separated list | none>
  Existing PSL assignment .......... <yes | no | n/a>
  Existing permset assignment ...... <yes | no>
  Write PSL ........................ <succeeded | already-had | skipped | FAILED>
  Write permset .................... <succeeded | already-had | skipped | FAILED>
  Verify actions surface ........... <yes | partial | no>

Verdict: ASSIGNED | ALREADY-ASSIGNED | HAND-OFF | VERIFY-INCONCLUSIVE | FAILED
Reason:  <plain-language explanation, or empty on success>

Next steps:
  - <If ASSIGNED / ALREADY-ASSIGNED: "Re-run service-itsm-agentic-setup-fulfiller-agent-configure — the invocable actions should now surface, and the activate call will succeed.">
  - <If HAND-OFF: "No Fulfiller persona is provisioned on this org. Run service-itsm-agentic-setup-agentforce-studio-validate to identify which AddOn needs enabling.">
  - <If VERIFY-INCONCLUSIVE: list the observed state verbatim; a session refresh or a different persona may be required>
  - <If FAILED: list the observed error verbatim + remediation>

Keep internal jargon (record Ids, HTTP status codes, FUNCTIONALITY_NOT_ENABLED, DUPLICATE_VALUE, object/developer names, sf api request rest) out of user-facing output.


Reference File Index

FileWhen to read
references/permset-topology.mdAny change to the persona list — the three Core-shipped Fulfiller permsets, their AddOns/PSLs/userPerms, and the fixed-lookup discovery query
references/cli-invocation.mdEvery phase — exact sf api request rest / sf data query / sf org assign permset call shapes, the never-extract-token rule, response envelopes
references/helper-contracts.mdThe input/output shapes of all four helper scripts (classify-permset-availability.mjs, resolve-target-user.mjs, classify-assignment-state.mjs, classify-action-surface.mjs) and how to interpret each verdict

FAQ

What does the service-itsm-agentic-setup-itsm-agentforce-permset-assign skill do?

Resolve missing ITSM Intelligence invocable actions so a Fulfiller NGA agent can activate. Reads which of the three Core Fulfiller persona permsets (IncidentFulfiller, ProblemFulfillerPermSet, ChangeRequestFulfillerPermSet) are provisioned, then assigns the running user the selected persona (plus backing PSL when license-gated) so svcitsmintelligence actions surface; hands off to service-itsm-agentic-setup-agentforce-studio-validate if none are provisioned. Use when the Fulfiller agent-configure skill reports missing actions on activate, when 'Invocable action svcitsmintelligenceX does not exist' surfaces, when a user asks to grant themselves Fulfiller prompt-template access, when asked to assign the Incident, Problem, or Change Fulfiller persona permission set, or when resolving missing ITSM Intelligence action access for the Fulfiller agent. DO NOT TRIGGER for Employee-agent access, Agentforce for IT Service toggles, agent creation, CMDB access, or generic permset assignment

How do I install service-itsm-agentic-setup-itsm-agentforce-permset-assign?

Run npx skills add https://github.com/forcedotcom/sf-skills --skill service-itsm-agentic-setup-itsm-agentforce-permset-assign in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/service-itsm-agentic-setup-itsm-agentforce-permset-assign/ (personal) or .claude/skills/service-itsm-agentic-setup-itsm-agentforce-permset-assign/ (project).

Which AI agents can use service-itsm-agentic-setup-itsm-agentforce-permset-assign?

Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.

Is service-itsm-agentic-setup-itsm-agentforce-permset-assign safe to install?

No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.

Who maintains service-itsm-agentic-setup-itsm-agentforce-permset-assign?

It is published by forcedotcom in forcedotcom/sf-skills. It has 1,176 recorded installs.

More from forcedotcom/sf-skills

View all →

platform-apex-generate

forcedotcom/sf-skills

Primary Apex authoring skill for class generation, refactoring, and review. ALWAYS ACTIVATE when the user mentions Apex, .cls, triggers, or asks to create/refactor a class (service, selector, domain, batch, queueable, schedulable, invocable, DTO, utility, interface, abstract, exception, REST resource). Use this skill for requests involving SObject CRUD, mapping collections, fetching related records, scheduled jobs, batch jobs, trigger design, @AuraEnabled controllers, @RestResource endpoints, custom REST APIs, or code review of existing Apex.

platform-apex-test-generate

forcedotcom/sf-skills

Use to generate and validate Apex test classes with TestDataFactory patterns, bulk testing (251+ records), mocking, assertions, and disciplined test-fix loops. Use when creating Apex test classes (for triggers, services, controllers, batch jobs, queueables, and callouts), improving coverage, debugging or fixing failing Apex tests, or running test/coverage analysis. Triggers on Test.cls files, sf apex run test, coverage reports. Do NOT trigger for production Apex (use platform-apex-generate) or Jest/LWC tests.

Testing & QA7.1K installs

platform-apex-test-run

forcedotcom/sf-skills

Apex test execution, coverage analysis, and test-fix loops with 120-point scoring. Use when the user runs Apex tests, checks code coverage, fixes failing tests, or touches Test.cls / Test.cls files. DO NOT TRIGGER when writing Apex production code (use platform-apex-generate), Agentforce agent testing (use agentforce-test), or Jest/LWC tests (use experience-lwc-generate).

Testing & QA7K installs

platform-apex-logs-debug

forcedotcom/sf-skills

Salesforce debug log analysis and troubleshooting with 100-point scoring. TRIGGER when: user analyzes debug logs, hits governor limits, reads stack traces, or touches .log files from Salesforce orgs. DO NOT TRIGGER when: running Apex tests (use platform-apex-test-run), generating or fixing Apex code (use platform-apex-generate), or Agentforce session tracing (use agentforce-observe).

platform-soql-query

forcedotcom/sf-skills

Use when the user needs SOQL/SOSL authoring or optimization: natural-language-to-query generation, relationship queries, aggregates, query-plan/selectivity analysis, and performance or safety improvements. TRIGGER when the user writes, optimizes, or debugs SOQL/SOSL, touches .soql files, or asks about relationship queries, aggregates, or query performance. DO NOT TRIGGER for bulk data operations (use platform-data-manage), Apex DML logic (use platform-apex-generate), or report/dashboard queries.

platform-metadata-deploy

forcedotcom/sf-skills

Salesforce DevOps automation using sf CLI v2. TRIGGER when: user deploys metadata, validates a deploy or does a dry-run, deploys to a Production org, quick-deploys a validated job ID, ships/promotes to prod, creates/manages scratch orgs or sandboxes, sets up CI/CD pipelines, or troubleshoots deployment errors with sf project deploy. DO NOT TRIGGER when: writing Apex code, building LWC components (use experience-lwc-generate), creating metadata definitions (use platform-custom-object-generate), deleting metadata from an org (use platform-destructive-deploy), or querying org data (use platform-data-manage).

DevOps & Cloud6.8K installs

Popular ai & agents skills

View category →

google-agents-cli-publish

google/agents-cli

This skill should be used when the user wants to "publish an agent", "publish my ADK agent", "register an agent with Gemini", "publish to Gemini" (formerly Gemini Enterprise), or needs guidance on the agents-cli publish gemini command. Also use when the user wants to "manage agents in Agent Registry", "list/update/delete registered agents", or "register an MCP server". Covers ADK vs A2A registration modes, programmatic and interactive usage, flag reference, auto-detection from deployment metadata, Agent Registry fleet management, and troubleshooting. Part of the agents-cli skills suite. Do NOT use for deployment (use google-agents-cli-deploy).

AI & Agents513.9K installs

google-agents-cli-eval

google/agents-cli

This skill should be used when the user wants to "run an evaluation", "evaluate my agent", "evaluate my ADK agent", "write an eval dataset", "analyze eval failures", "compare eval results", "optimize agent", or needs guidance on the Gemini platform eval methodology and the Quality Flywheel. Covers eval metrics, dataset schema, LLM-as-judge scoring, and common failure causes. Applies to any agents-cli project, whatever framework the agent is written in. Do NOT use for agent API code patterns (ADK: use google-agents-cli-adk-code), deployment (use google-agents-cli-deploy), or project scaffolding (use google-agents-cli-scaffold).

AI & Agents513.5K installs

skill-creator

anthropics/skills

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

OfficialAI & Agents416.4K installs

codex-pet

prime-skills/runcomfy-agent-skills

Codex Pet generator on RunComfy. Build a Codex-compatible Codex Pet spritesheet.webp + pet.json from a single reference image, drop it into ${CODEXHOME:-$HOME/.codex}/pets/ / and Codex picks it up as a custom Codex Pet next to the 8 built-ins. This skill produces the exact Codex Pet atlas Codex expects (1536x1872 PNG/WebP, 8 cols x 9 rows, 192x208 cells, 9 animation states — idle, running-right, running-left, waving, jumping, failed, waiting, running, review). Calls OpenAI GPT Image 2 edit ONCE via the local RunComfy CLI as runcomfy run openai/gpt-image-2/edit to produce a canonical Codex Pet pose, then assembles all 9 animation rows programmatically with ImageMagick micro-transforms — no Codex Pro, no $imagegen, no OPENAIAPIKEY required, only RUNCOMFYTOKEN. Triggers on "codex pet", "create codex pet", "make codex pet", "hatch codex pet", "/hatch image", "desktop pet codex", "codex pets", "spritesheet.webp", or any explicit ask to build a custom pet for OpenAI Codex.

AI & Agents402.1K installs

full-output-enforcement

leonxlnx/taste-skill

Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.

AI & Agents401K installs

executing-plans

obra/superpowers

Use when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available

AI & Agents245.7K installs