feat(ai-delegation): add delegate-to-router and openrouter-models skills - #450
Merged
openbao-service-broker[bot] merged 1 commit intoAug 2, 2026
Merged
Conversation
Two harness-agnostic skills for offloading bounded subtasks to a shared OpenAI-compatible model router. delegate-to-router covers the mechanics: read the endpoint from the environment, fetch the served model menu from the router's own contract rather than remembering a name, bound every call with an explicit timeout, and handle each failure mode as a stated outcome instead of a silent fallback into the caller's own context. openrouter-models covers selection and cost: discover current ids, prices, and context lengths from the public keyless catalog, self-enforce a daily spend budget where nothing meters it for the caller, keep confidential material off prompt-logging free tiers, and route a request for an unserved model through infrastructure rather than self-service. Neither skill names a model id. Model inventories move faster than documents do, so a name written here would be a second spelling that drifts from the router's registry — the duplication these skills exist to remove. Both are written against shell, curl, and jq only, so non-Claude harnesses consume them from here instead of maintaining a second authored copy. Frontmatter keeps version and author under metadata: the agentskills.io spec this repo validates against allows neither at the top level.
This was referenced Aug 2, 2026
openbao-service-broker
Bot
added
the
human:review
openbao-service-broker
Bot
deleted the
feat/2c90b02d-router-delegation-skills
branch
August 2, 2026 19:09
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds two skills to the
ai-delegationplugin:delegate-to-router— the mechanics of offloading a bounded subtask to ashared OpenAI-compatible model router: read the endpoint from the environment,
fetch the served model menu from the router's own contract, bound the call, and
handle each failure mode explicitly.
openrouter-models— selection and cost: discover current ids, prices, andcontext lengths from the public keyless catalog, self-enforce a daily spend
budget, keep confidential material off prompt-logging free tiers, and route a
request for an unserved model through infrastructure rather than self-service.
Why here
These were first written in
ai-assistant-instructions, which states it "shipsconfiguration only — no slash commands, skills, agents, or hooks". Putting skills
there broke that boundary, so they move to the marketplace, which is where every
skill in this estate already lives. The rule that governs them stays in
ai-assistant-instructions(dryvist/ai-assistant-instructions#793) — a rule isconfiguration; the procedure is a skill.
Single authored copy
Both skills are harness-agnostic on purpose: shell,
curl, andjqonly, nomodel ids, endpoint read from the environment. That lets non-Claude consumers pin
this repository as a plain source input and take the same file rather than
maintaining a parallel copy. dryvist/nix-hermes#43 does exactly that, and deletes
its local
openrouter-modelsin the same change — two copies of a skill aboutspend and egress would drift on precisely the rules that carry the risk.
Frontmatter note
versionandauthorsit undermetadata:rather than at the top level. Theagentskills.iospec this repo validates against (validate-skills.yml) allowsonly
name,description,license,metadata,allowed-tools, andcompatibilityat the top level, and rejects YAML flow sequences in frontmatter.The skill bodies are byte-identical to the reviewed originals.
Verification
Run locally against this branch:
agentskills validate— all fiveai-delegationskills passmarkdownlint-cli2— 0 errors acrossai-delegation/**andproject-standards/**plugin.jsonskills entries match the schema's^\./skills/[a-z0-9-]+$patternVersions are left alone — release-please owns them.