feat(dsh): add verified skill discovery - #100
Merged
Merged
Conversation
Define Story QoderAI#99 as the DSH move from Partial adapter toward Verified install/discovery. The native customSkillDirs route and credential-free evidence boundary are fixed while the cross-host-safe explicit-only mechanism remains implementation research. Validated with the repository Spec Preparation contract, structured specification review, documentation link tests, and attribution gates. Co-authored-by: Codex (GPT 5.6 Sol) <codex@openai.com>
Define the canonical winner, trust, path, and explicit-only behavior that the DSH integration must satisfy. The tests were observed failing before the DSH policy owner existed. Co-authored-by: Codex (GPT 5.6 Sol) <codex@openai.com>
Add a DSH-native Cordis policy that verifies the winning Better Harness Skill against the complete configured root before direct slash invocation and rejects model-facing Skill tool calls. Add a pinned credential-free native smoke to CI. Co-authored-by: Codex (GPT 5.6 Sol) <codex@openai.com>
Exercise the real DSH scoped registry used by selected Web presets, while retaining the unsupported no-loader boundary represented by Web minimal. Co-authored-by: Codex (GPT 5.6 Sol) <codex@openai.com>
Document the single canonical customSkillDirs route, the headless and supported Web preset owners, explicit-only policy, trust failures, path rules, and the capabilities that remain unsupported. Co-authored-by: Codex (GPT 5.6 Sol) <codex@openai.com>
Keep the DSH installation reference as prose so the Antigravity Markdown artifact closure remains byte-for-byte equivalent to the current upstream/main closure rather than absorbing unrelated site pages. Co-authored-by: Codex (GPT 5.6 Sol) <codex@openai.com>
Run the credential-free DSH owner smoke immediately after install so its macOS, Linux, Windows, and Node matrix evidence is reported independently of later repository-wide gates. Co-authored-by: Codex (GPT 5.6 Sol) <codex@openai.com>
Run npm through its JavaScript CLI entrypoint when the smoke is started by npm, avoiding Node 22.20 spawn EINVAL for npm.cmd on Windows while retaining a shell fallback for direct execution. Co-authored-by: Codex (GPT 5.6 Sol) <codex@openai.com>
Align the traceability status with the completed Story QoderAI#99 acceptance evidence before PR QoderAI#100 is merged. Validated with the focused DSH and documentation link tests. Co-authored-by: Codex (GPT 5.6 Sol) <codex@openai.com>
1 task
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
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
Adds a verified DeepSeek Harness (DSH) installation and discovery path for the canonical Better Harness Skill, with direct-user-only
/better-harnessinvocation enforced by a DSH host policy.Why
Traceability and Scope
docs/specs/2026-08-22-99-deepseek-harness-skill-discovery.mdChange Type
Test and Review Evidence
npm run test:dsh-native0.1.1-rc.2, credential-free native owners, global Headless and scoped Web discovery, explicit injection, model rejection, shadows/fallbacks, and path semantics verifiednpm test -- --run test/skills-docs/doc-link-graph.test.mjs test/skills-docs/dsh-skill-discovery.test.mjs test/skills-docs/better-harness-skill.test.mjsnpm run harness:generatednpm run pack:verifynpm --prefix packages/harness-studio run test:browsernpm run buildgit diff --checkmain: 101/102 files pass, 1482 passed, 1 skipped; the two failures are pre-existing stale Antigravity closure counts (expected 93/290/96, actual 102/299/105) and reproduce at upstream SHA465e9bd/ CI run3255293345232556739754mainAntigravity assertions in the shared Vitest stepManual or visual evidence: reviewed the complete policy owner, native smoke, tests, docs, workflow diff, package diff, commit trace, and current upstream CI logs. The branch preserves the same closure graph as
upstream/main(102 nodes, 299 edges, 105 files).Risk and Recovery
~is not expanded.AI Involvement
Checklist
AGENTS.md,CONTRIBUTING.md, and the relevant canonical-owner guidance.CHANGELOG.md(explicit Issue [Feature]: Add verified DeepSeek Harness Skill discovery with explicit-only invocation #99 non-goal; canonical adapter and installation docs are updated instead).