engine: 028 face 5 — declared-but-empty vault key warns, never blocks - #4
Open
Morgandri1 wants to merge 3 commits into
Open
engine: 028 face 5 — declared-but-empty vault key warns, never blocks#4Morgandri1 wants to merge 3 commits into
Morgandri1 wants to merge 3 commits into
Conversation
|
Deployment failed for project wheel with the following error: Learn More: https://vercel.com/morgandri1s-projects?upgradeToPro=build-rate-limit |
|
The latest updates on your projects. Learn more about Vercel for GitHub. |
Morgandri1
force-pushed
the
sdk/028-face5-declared-overlap-warning
branch
from
September 6, 2026 16:52
2459a0d to
2824759
Compare
added 3 commits
September 6, 2026 17:46
find_ambiguity judged presence on offered_keys (declared ∪ stored), so a vault that merely DECLARED a credential key with no value stored still 409-blocked wiring a second vault that actually held it. PM overruled this: presence is a stored value everywhere (face 1 already fixed `GET .../auth` on this basis); a declaration is a statement of intent, not a competing credential. find_ambiguity is now stored-only (list_keys) and is the only thing that blocks a wire or an agent's env. The declared/declared case is still worth surfacing — two vaults both intending to supply the same credential is real information for the operator — so it is now find_declared_overlap, returned as a non-blocking `warning` in POST /v1/wires's response instead of a 409. Acceptance: redteam/pocs/vault/run_declared_empty.sh §5 (v declares CLAUDE_CODE_OAUTH_TOKEN with no value; v2 holds the real one; wiring the agent to v2 must succeed, not 409).
Written from a fresh wheel-dev session so the next stand-down (or the next session of me) has this turn's state without re-deriving it.
…ck a real one there either
PM ruling: the wire-time fix (find_ambiguity/find_declared_overlap) and
the PUT-time check (supplies_key, in store_in_vault_until) are the same
finding class -- presence is a stored value everywhere. supplies_key
used offered_keys (declared union stored), so a vault that only
DECLARED a key still 409-blocked PUTting the real value into a second
vault the same agent reads.
supplies_key is now stored-only (list_keys) -- the only thing that
blocks a PUT. A new declares_key (declared-only) surfaces the
declared/declared overlap as a non-blocking `warning` in the response,
same shape as add_wire's: PUT /v1/vault/:id/:key now returns
{"key","stored":true,"warning"?} instead of bare {"key","stored":true},
and the paste-code login route (agent_routes.rs) merges this into its
existing warning field rather than clobbering the session-credential
one.
Morgandri1
force-pushed
the
sdk/028-face5-declared-overlap-warning
branch
from
September 6, 2026 17:47
ccd0a90 to
c8372f1
Compare
Morgandri1
added a commit
that referenced
this pull request
Sep 6, 2026
The 202 envelope's `delivered` counts rows the engine ENQUEUED (wheel-engine/src/api/ingress.rs:331-345 increments per successful messages::enqueue, and discards the spawned supervisor.start result), so a parked, stopped or unauthenticated agent still returns delivered:1. Under the message states (§3c #4) `delivered` means the bytes reached the child's stdin. The panel was repeating the wire's word and adding emphasis to it — "This was a real hit, not a simulation" — which is the strongest possible claim on the weakest available evidence, aimed at an operator who is testing precisely because they do not yet trust the path. It now reports the count and says queued, naming the parked case. The test asserts the number survives AND that the sentence cannot claim receipt; the old wording fails it.
Morgandri1
added a commit
that referenced
this pull request
Sep 7, 2026
Was pre-M1 (findings 001-011, all design-review). Now: stack boots+deployed, most findings CONFIRMED-by-run or CLOSED. Leads with PM's theme — the convincing half is built, the load-bearing half is not (048 network, 047 re-check, 037 per-node), and the measure-vs-claim discipline incl. my own 048 correction. Open/load-bearing table, closed highlights, top-3 (gate script-exec on 037/038 + 048; land per-node uids; deploy S5b segmentation + deploy-check). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TbyG7euRsVotMStegDRxRY
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.
Summary
crates/wheel-engine/src/vault.rs: splitfind_ambiguity(wasoffered_keys= declared ∪ stored) into a stored-onlyfind_ambiguity(the only thing that blocks a wire or an agent's env — a real credential clash) and a newfind_declared_overlap(declared-only, non-blocking).crates/wheel-engine/src/db/board.rs:add_wirestill refuses a STORED clash (409ambiguous_credential), but now returnsOk(Some(warning))instead of erroring when two vaults merely declare the same key.crates/wheel-engine/src/api/board_routes.rs:POST /v1/wiresreturns200 {}normally,200 {"warning": "..."}when a declared overlap exists — never a 409 for that case. (Web'srequest()helper already handles a JSON 200 the same as the old 204; the one caller,canvas.tsx, doesn't read fields off the response, so this is compatible today.)a_declared_but_empty_key_does_not_block_the_vault_with_the_real_value) plus one proving a real stored clash still blocks (two_stored_values_for_the_same_key_are_still_blocked); fixed two pre-existing tests that relied on declared-only ambiguity to store real values instead.Ref:
redteam/findings/028-declared-but-empty-credential-reports-authenticated.mdface 5, PM's overrule indocs/handoff/sdk.mditem 4.Test plan
cargo build --workspacecargo test -p wheel-engine— 265 passedcargo clippy -p wheel-engine --all-targets -- -D warnings— cleancargo fmt -p wheel-engine -- --check— cleanredteam/pocs/vault/run_declared_empty.sh§5 live against a rebuilt image (no docker in this sandbox to run it here) — asking ADVERSARY to re-verify per their handoff doc