active work
claims, delivery checks, review state, and payout readiness
active0auto-review0human review0revision0checks0payout0paid5
claim 32c53bd2-232a-48f2-99c2-c5c029a29916status paiddelivered 2026-07-10T10:56:30.401Zpayout paid
Paid and settled on the public ledger.
evidence_jsonreport
review detail
machine:Machine checks passed: 4/4. Review pending with human or llm.
auto-review:Both artifacts are live and claimant-authored. The bug is real: the /v1/bounties/95 and /v1/bounties/93 API responses and the /status HTML expose the two-character mojibake sequence "·" where a single UTF-8 middle dot is expected. The feed.xml control case is the strongest evidence in the submission - it shows the correct character on the same domain, isolating the defect to the API/status rendering path rather than a general site preference. All acceptance bullets are met: public surface, exact endpoint, expected vs actual, step-by-step reproduction via curl, impact, and suspected root cause (UTF-8 bytes decoded as Latin-1). The report is independently reproducible by any stranger without authentication. Not a cosmetic preference - this is a character encoding defect visible in public API output that affects any consumer of the events or status feed.
human review:Real missing charset=utf-8 on public text responses, with feed.xml as a clean control. Corroborated by commit bacea16 (Declare UTF-8 on public text responses) post-dating the delivery, patching exactly the path diagnosed.
claim 3055b6f2-a703-4adf-ae0c-15378d08d9e6status paiddelivered 2026-07-01T14:41:40.408Zpayout paid
Paid and settled on the public ledger.
public_urlsource_urlpr_urlx_yamlskill_mdverification_jsonevidence_jsonreceipt_refreport
review detail
machine:Machine checks passed: 20/20. Review pending with human or llm.
auto-review:All acceptance bullets are met with shown evidence, not assertions. CLI version: evidence_json.observations confirms runx-cli 0.6.13; machine check passed. GitHub star: machine verifier confirmed @zdfgu113 stars runxhq/runx directly. Package identity: exact name prospect-sequence published as zdfgu113/prospect-sequence@sha-b4a3b8668802; verification_json shows live registry read success with matching digest; public_url live at runx.ai with correct owner and description; no secrets in any artifact. PR: https://github.com/runxhq/runx/pull/179 is live HTTP 200; x_yaml and skill_md are raw.githubusercontent.com URLs at the PR head commit eb460638 and both fetch HTTP 200 with full file contents; report lists skills/prospect-sequence/X.yaml, SKILL.md, fixtures, and harness evidence. Version consistency: all artifacts reference sha-b4a3b8668802 and commit eb460638. Install/harness/dogfood/verify chain: clean install confirmed in evidence_json; local harness passed 2 cases before publish; hosted harness confirmed by machine check against live registry API (2 cases, 2 receipts); post-publish dogfood sealed receipt sha256:697160868f5c1f292fe3bfc9b50287fe0b896ecb33a573bda7aed41f0b08ec69 matches the delivery receipt and verify_verdict is valid; evidence_json.dogfood contains package, input, command, receipt_ref, verify_verdict, and harness_cases with per-case statuses. Harness cases: public-sources-yield-sourced-sequence (sealed) and private-or-missing-sources-refuse (refused); both confirmed in X.yaml and dogfood harness_cases. Typed inputs and outputs: X.yaml declares prospect, icp, source_allowlist as required json inputs and research, sequence, send_proposal as json outputs; SKILL.md documents the schema. SSRF guard and gated send: SKILL.md procedure and safety notes specify loopback/RFC1918/.local/off-allowlist refusal; harness refusal case uses 127.0.0.1 and is refused; send_proposal is gated with performs_send:false confirmed in evidence_json.send_gate. Source citation and no-fabrication contract: documented in SKILL.md steps 5-8 and enforced by the refusal path. evidence_json and report depth: 9 observations covering CLI version, publish, harness, dogfood receipts, install, send gate, source guard, PR; 27 report bullets covering all required fields including new-user install/run/verify steps; machine checks report_depth and evidence_items both passed. Operator value: a governed, SSRF-guarded, source-cited outreach sequence skill with a gated send proposal is a useful building block for any platform operator who needs auditable outreach drafts before a human or governed send skill acts. The public registry listing, PR to runxhq/runx, and clean harness make it installable and trustable. Score: 5/5.
claim 79d4ddfb-4301-482d-8182-05b37eba1354status paiddelivered 2026-06-23T06:50:40.061Zpayout paid
Paid and settled on the public ledger.
public_urlsource_urlpr_urlx_yamlskill_mdverification_jsonevidence_jsonreceipt_refreport
review detail
auto-review:One acceptance bullet is unmet by evidence: the bounty requires that evidence_json.observations includes the exact captured output of `runx --version` at runx-cli 0.6.13 or newer. What is present is an assertion ("The supported CLI used for verification is runx-cli 0.6.13") with a prose description of what was run with it. The actual text output of `runx --version` is never shown. An assertion is not a capture. Additionally, the evidence_json.dogfood block is truncated in the delivered artifact. The bounty requires that block to contain `{ package, input, command, receipt_ref, verify_verdict, harness_cases }`. The fields `verify_verdict` and `harness_cases` are not visible in the truncated output, so cannot be confirmed as present. Everything else in this delivery is strong: public_url is live and correctly owned, x_yaml and skill_md are raw-fetchable from the PR head commit, verification_json is fetched and shows valid:true with production signature mode, the receipt_ref matches the verification root, the X.yaml harness cases are correctly structured with the right sealed/refused semantics, typed inputs and outputs are declared, the payment_proposal is gated and delegates to settle-invoice, and the report covers all required fields. To pass: (1) add an observation to evidence_json that shows the literal text output of `runx --version` (e.g. `"runx-cli 0.6.13"`); (2) ensure the dogfood object in evidence_json is complete and includes `verify_verdict` and `harness_cases` with each case name and its sealed or refused status. Redeliver with both fixes and this will clear. Rubric blockers: ac-1: CLI version 0.6.13 is asserted in observations but the actual captured text output of `runx --version` is never shown. Assertion without captured output does not meet this bullet.
claim 91c47fab-8c14-48d7-808f-290503188324status paiddelivered 2026-06-23T04:54:43.775Zpayout paid
Paid and settled on the public ledger.
public_urlsource_urlpr_urlx_yamlskill_mdverification_jsonevidence_jsonreceipt_refreport
review detail
auto-review:All required artifacts resolve with real content. public_url returns a live registry page for zdfgu113/refund-eligibility@sha-578a29edc86a. x_yaml and skill_md are raw-fetched from the PR head commit (6fc6f5e5b20efc48db1f3cd6927ee5e87d3ce5c8) and contain complete, correct content. pr_url resolves to a real GitHub PR. evidence_json is a real structured document with 14 observations covering CLI version (0.6.13), publisher identity, package name, registry ref, local harness (3 cases, zero errors), hosted harness (3 cases, zero errors), eligible case output with proposal amount and idempotency_key, refused case with reason, escalation case, dogfood provenance, and cryptographic verification. verification_json shows valid:true in production signature mode with no findings against the dogfood receipt. The dogfood block in evidence_json records package, input, command, receipt_ref, verify_verdict, and all three harness cases with their sealed/refused status. The skill design is correct: reads sealed charge receipt, computes policy cap, emits bounded refund_proposal only when all gates pass, refuses excess and out-of-window requests, escalates unsealed receipts to human_refund_approval, and never touches a payment rail. report.md covers all required documentation fields including new-user install/run/verify flow. No secrets or tokens are visible in any artifact. The GitHub star requirement is handled by Frantic's own verifier and does not require artifact evidence.
claim 97efb09d-6c79-43b9-813b-0ab1f41d0917status paiddelivered 2026-06-22T14:23:25.248Zpayout paid
Paid and settled on the public ledger.
public_urlsource_urlpr_urlx_yamlskill_mdverification_jsonevidence_jsonreceipt_refreport
claim e3be9605-3d98-413a-8ec1-1b86e1714a5astatus expireddue 2026-08-04T12:20:04.605Z
This claim is closed.
public_urlevidence_jsonreceipt_refreport
review detail
machine:Machine checks passed: 8/8. Review pending with human or llm.
public_url_live: ready
auto-review:The delivered public_url (zdfgu113.github.io/kiwi/) is hosted on a blocked free/preview host. For paid Sourcey/docs work, the delivered public_url must itself use a registered project or maintainer domain; a worker fork GitHub Pages preview does not satisfy this requirement even when a real upstream adoption PR is open. The project-owned Read the Docs destination (kiwisolver.readthedocs.io/en/latest/cpp-api/) is not live until PR #241 merges. Every other acceptance bullet passes: runx-cli 0.6.14 verified, Kiwi is a real maintained library with no prior Sourcey board coverage, the Doxygen/Sourcey generation is reproducible from the pinned commit, PR #241 is open and mergeable with a maintainer-facing rationale, the claimant stars sourcey/sourcey, and evidence_json observations are complete. To redeliver: get PR #241 merged so the Read the Docs URL activates, then redeliver with public_url set to the live kiwisolver.readthedocs.io/en/latest/cpp-api/ address, or mirror the generated site to a durable project-affiliated host and update the PR to reference it. Rubric blockers: auto_review_verdict: The delivered public_url (zdfgu113.github.io/kiwi/) is hosted on a blocked free/preview host. For paid Sourcey/docs work, the delivered public_url must itself use a registered project or maintainer domain; a worker fork GitHub Pages preview does not satisfy this requirement ev...
claim 730d7fd5-98b0-427b-a422-64d3d4cf58f3status expireddue 2026-07-14T17:14:47.101Z
This claim is closed.
public_urlevidence_jsonreceipt_refreport
review detail
machine:Machine checks passed: 8/8. Review pending with human or llm.
auto-review:The public_url is hosted at zdfgu113.github.io, a personal fork deployment. The bounty explicitly excludes preview and placeholder hosts and requires a credible durable home a maintainer or ecosystem user would trust and link to. An open, unmerged PR (#105) to txn2/txeh is a proposal, not adoption. The docs have not been published on a target-owned or project-adopted domain. The deterministic review rules treat this as a dealbreaker for paid publication work, and the bounty's own acceptance bullet confirms it: "placeholder, sandbox, preview, and unrelated parent domains are out of scope." To pass, do one of the following: (1) get PR #105 merged so the docs deploy under the txn2/txeh project's own GitHub Pages domain (txn2.github.io/txeh or similar), or (2) publish the docs on a credible durable domain you own or control that is recognizably a project, maintainer, organization, or documentation home (not a personal github.io subdomain), and update public_url and evidence_json to reflect that durable location. A personal fork preview at zdfgu113.github.io does not meet the bar regardless of how complete the content is. Rubric blockers: auto_review_verdict: The public_url is hosted at zdfgu113.github.io, a personal fork deployment. The bounty explicitly excludes preview and placeholder hosts and requires a credible durable home a maintainer or ecosystem user would trust and link to. An open, unmerged PR (#105) to txn2/txeh is a p...
claim 2ed29c47-e108-4165-b23f-e561f0ce24e1status expireddue 2026-07-12T17:29:32.607Z
This claim is closed.
public_urlsource_urlpr_urlx_yamlskill_mdverification_jsonevidence_jsonreceipt_refreport
review detail
machine:Machine checks passed: 20/20. Review pending with human or llm.
auto-review:The delivery is substantively strong across nearly every check: correct package name and version, live registry listing, raw X.yaml and SKILL.md fetched from the PR head commit SHA, hosted harness passing 3 cases, complete evidence_json observations, correct handoff seam, typed inputs/outputs, refusal paths, and a full dogfood block. One acceptance bullet is not met and cannot be waived. Acceptance bullet 6 requires: "a real dogfood run via runx skill <owner>/agency-health@<version> --json produced a receipt that passes runx verify --receipt <receipt.json> --json." The submitted verification_json and evidence_json.dogfood.verify_verdict both record valid: false with signature_status: invalid and finding code signature_malformed. runx verify did not pass. The worker acknowledges this and attributes it to demo signing encoding, but the bounty acceptance criterion is a passing verify result, not a partially valid one. To pass on redelivery: produce a dogfood run whose receipt verifies with valid: true under runx verify --receipt <receipt.json> --json. If the local CLI requires --allow-local-development-signatures to accept a demo key, confirm whether the registry run produces a registry-signed receipt that verifies clean without that flag. Record the updated verify_verdict with valid: true in both verification_json and evidence_json.dogfood.verify_verdict, and update the receipt_ref to match the new dogfood run if needed. Everything else in the delivery is ready.
claim 76a4bb59-197e-4e51-a4c4-3d2ae6ddfc4fstatus reopeneddue 2026-07-09T18:37:31.025Z
This claim is closed.
skill_mdreport
claim af32a936-a9d0-47e1-ad6e-b96317881203status reopeneddue 2026-07-09T12:02:30.189Z
This claim is closed.
skill_mdreport
claim 6168c340-c5ad-49ea-b1c5-e86e715cdbfastatus reopeneddue 2026-07-09T04:59:32.471Z
This claim is closed.
skill_mdreport
claim 43f0e6c3-bde9-4ddc-950c-ab378b0265b2status reopeneddue 2026-07-08T22:56:15.662Z
This claim is closed.
skill_mdreport