Name the identity fields oda_map_projects requires - #59
Merged
Conversation
`fields` on oda_map_projects looks like a free choice among fourteen values, but every returned item must carry id and name, and a gateway that has not taken the identity fix blocks any array missing either one with PUBLIC_RESPONSE_BLOCKED. The Skills walked consumers into that. "Selecting fields" told them to request only what a question needs and never mentioned the two required values, and the blocked-response guidance told them to narrow the request — which, for this failure, keeps it blocked. A lookup that hit this dropped its agency filter, paged the whole country list, and tallied by hand, having misread the block as a filter problem. Records the requirement where fields are chosen, and puts the fields check ahead of narrowing wherever a blocked response is handled. The gateway fix (amnotyoung/io-mcp) fills id and name in, so the wording stays conditional on the deployed build rather than claiming the block is permanent. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
amnotyoung
added a commit
that referenced
this pull request
Aug 13, 2026
#59 told readers to put id and name in every `fields` array because the gateway blocked any array without them. amnotyoung/io-mcp#61 shipped, so the gateway fills both in and nothing blocks. Instructions written against a defect outlive it badly: they read as requirements, and the next person spends time satisfying one that no longer exists. Restores the three Skill files to their pre-#59 wording — the blocked-response sections lose the fields step they gained, and the callers that named id and name alongside description drop them again. Keeps one line, because one thing #59 found is still true and was never written down: every item carries id and name, so they come back whether or not the caller lists them. That is why the response echo can be longer than the request, which is otherwise a surprise with no explanation anywhere. The empty-projection fix (amnotyoung/io-mcp#62) needed no Skill change: it removed a false claim from responses rather than adding a rule for callers to follow. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
amnotyoung
added a commit
that referenced
this pull request
Aug 13, 2026
#59 told readers to put id and name in every `fields` array because the gateway blocked any array without them. amnotyoung/io-mcp#61 shipped, so the gateway fills both in and nothing blocks. Instructions written against a defect outlive it badly: they read as requirements, and the next person spends time satisfying one that no longer exists. Restores the three Skill files to their pre-#59 wording — the blocked-response sections lose the fields step they gained, and the callers that named id and name alongside description drop them again. Keeps one line, because one thing #59 found is still true and was never written down: every item carries id and name, so they come back whether or not the caller lists them. That is why the response echo can be longer than the request, which is otherwise a surprise with no explanation anywhere. The empty-projection fix (amnotyoung/io-mcp#62) needed no Skill change: it removed a false claim from responses rather than adding a rule for callers to follow. Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
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 happened
A plugin test run asked what KOICA funds in Tanzania. Five consecutive
oda_map_projectscalls came backPUBLIC_RESPONSE_BLOCKED; the sixth, which dropped bothagencyandfields, worked. The caller concludedagencywas at fault, abandoned the filter, paged all 177 country records, and tallied KOICA by hand.agencywas never the problem.{country, agency: "한국국제협력단(KOICA)"}returns 102 items, and addingstatus: "active"returns 49 — the exact number the hand tally reached before discarding it as untrustworthy. Every failed call omittedidfromfields.Why the Skills made it worse
"Selecting fields" told readers to request only what a question needs and never mentioned that each item must carry
idandname. The first blocked call —fields: ["name","sector","dates","status","description"]— was that guidance followed exactly.The blocked-response guidance then said to narrow the request. For this failure narrowing cannot help: a smaller array still omits the required values. That is why five attempts, shrinking
limit100 → 30 → 10 andfieldsfive → one, all failed.Change
references/portfolio-lookup-protocol.md— states the requirement as the first item under "Selecting fields", where fields are chosenfieldsbefore narrowingkorean-oda-portfolio-lookup/SKILL.mdstep 4 and the comparison workflow,international-oda-data-lookup/SKILL.mdroute 2 — nameidandnamealongside the fields they already requestgenerate-development-country-reportis untouched: it is a synced build output fingerprinted bypublic-skill.lock.json, and it carries nofieldsguidance to correct.Version 0.2.27 → 0.2.28;
npm test27/27.Relationship to the gateway fix
amnotyoung/io-mcp#61 fills
idandnamein, so the constraint disappears once deployed. The wording here is conditional on the deployed build rather than claiming the block is permanent, and these paragraphs can be dropped after that ships.🤖 Generated with Claude Code