Skip to content

Releases: aiopshwang/goal-to-proof

Release list

v1.1.0 — efps merged

v1.1.0 — efps merged Latest
Latest

Choose a tag to compare

View all tags
@aiopshwang aiopshwang released this 24 Aug 14:50
Immutable release. Only release title and notes can be modified.
v1.1.0
1a145ed

Merges the evidence-first-problem-solving diagnosis loop, hardens the completion gate with a rationalization table, and preserves efps eval provenance. See CHANGELOG.md.

Note: the forward-test record's permission-envelope wording was corrected on main in 4cd1711 after this tag was cut; the tagged snapshot carries the pre-correction wording.

Assets 3
Loading

Goal to Proof v1.0.0

Goal to Proof v1.0.0

Choose a tag to compare

View all tags
@aiopshwang aiopshwang released this 23 Aug 11:32
v1.0.0
879a47d

Goal to Proof is a lightweight closure contract for AI agents: finish authorized, non-trivial work and prove the requested outcome with direct, scope-matched evidence.

Install

Agent Skills:

npx skills add aiopshwang/goal-to-proof

Codex:

codex plugin marketplace add aiopshwang/goal-to-proof
codex plugin add goal-to-proof@goal-to-proof

Claude Code:

claude plugin marketplace add aiopshwang/goal-to-proof
claude plugin install goal-to-proof@goal-to-proof

Included

  • A portable Agent Skills package and native Codex and Claude Code marketplace manifests.
  • A narrow Result → Target → Proof → Boundaries completion contract.
  • 24 trigger and behavior cases, including five fail-closed hard gates.
  • A deterministic release validator, unit tests, security policy, contribution workflow, and public documentation site.
  • English and Korean documentation.

Verification for this release

  • Exact release commit passed 1,239 deterministic checks with zero errors and zero warnings.
  • All 24 repository tests passed.
  • The package passed the Agent Skills reference validator, Codex plugin validator, and Claude Code manifest and marketplace validators.
  • Install and explicit invocation were exercised from the exact committed tree with skills 1.5.23, Codex CLI 0.149.0-alpha.4.1, Claude Code 2.1.153, Python 3.14.5, and Node 25.9.0.
  • Selected positive, negative-trigger, false-completion, and target-boundary behavior cases were exercised live. These are scoped release checks, not a claim of universal performance improvement.

The source methodology was derived from aggregate patterns in prior real working sessions. Raw conversations, private corpus metadata, secrets, personal names, and one-off preferences are not included in the release.

Full documentation: https://aiopshwang.github.io/goal-to-proof/

Loading