feat(skills): decide — the free typed-judgment endpoint as a skill, not a tool - #152
Merged
Merged
Conversation
…aude Code, without an MCP tool POST api.blockrun.ai/v1/decide (OpenJev) is free behind a registered key and not x402. It is deliberately NOT a blockrun_decide tool: every MCP caller is already a frontier model and a better one-off judge than a 4B NLI cross-encoder, the default wallet install has no brk_ key, and a tool taxes every user's context on every turn for an experiment few run. The skill costs nothing until triggered and carries the whole recipe: curl, the three question types and their limits, the response shape, and the eval loop over the user's own labelled rows. Two things every surface has to say, and this one does: the `confidence` on a choice is a share of agreement, not a probability (equal-weak and equal-strong support are indistinguishable after normalisation), and OpenJev is not Jev — unaffiliated, MIT, no published comparison, none invented. Also tells the agent to export the key in the shell rather than write ~/.blockrun/.api-key, because that file flips every paid tool to account billing at the next start. Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
…stion at `decide` Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
A
decideskill forPOST api.blockrun.ai/v1/decide(OpenJev typed judgments), registered in the marketplace, plus the skill count (16 → 17) and a pointer from theblockrunrouter skill.Why a skill and not a
blockrun_decidetoolbrk_key; a tool would mostly return "go register".decide's real audience is programs without a model (SDK/API), which the docs page already serves. The skill exists for the "try it on my data from Claude Code first" case, viacurl.What the skill says that it must say
confidenceon a choice is a share of agreement found, not a probability the answer is correct (the equal-weak / equal-strong example from the API reference).~/.blockrun/.api-keyjust to try this, because that file moves every paid tool to account billing.Verification
npm test: 1312/1312 (marketplace reachability, frontmatter YAML, doc-file-refs, skills CLI all cover the new directory).npx tsx src/index.ts skills listshowsdecide.claude plugin validate .passes (pre-existing warning about a missing marketplace description only).invalid_api_keywith no key) and https://blockrun.ai/docs/api-reference/decide; attribution facts from https://blockrun.ai/openjev.No source, tool count, schema-token or CHANGELOG changes — nothing to release; the skill ships with the next version.
🤖 Generated with Claude Code