feat: narrow install surface to adapted agents + Hermes support - #6
Merged
Conversation
Restrict visible/installable tool adapters to agents with real Rasen support instead of listing every detected AI tool, so init/update no longer offer installs for tools that have no working adapter. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…a2a52ce) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Wire Hermes into the adapter surface alongside Claude and Codex, with its own global skills-home resolution, config, migration, and profile-sync-drift handling. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…egated to children, both children shipped+archived) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
DumoeDss
added a commit
that referenced
this pull request
Jul 16, 2026
Resolves both conflicts: - test/core/init.test.ts: take dev/0.1.4's adapted-tools semantics for --tools all (agent-adapters PR #6) - test/core/templates/skill-templates-parity.test.ts: recompute both SHA256 snapshot maps from the merged templates (12 entries — the union of both sides' template ripple sets), via the test's own hashing method Full suite on merged tree: 2799 passed / 0 failed (after dist rebuild). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K6AuE3xd23DEeikmwXQtYx
DumoeDss
added a commit
that referenced
this pull request
Jul 28, 2026
… wrapping, test fixes Resolves the 6 remaining CI failures on PR #88 head (e3532d0): - #2 audits-api socket hang: rejectImport uses socket.end() (FIN) + destroy fallback, not req.destroy() (RST) - #4 B3 race loser: snapshot fs.existsSync before probe; absent-at-start + later-refused = obtain-failed (lost race) - #5 token-audit: read() wraps all errors as AuditServiceError (was ELOOP on Linux symlinks) - #1 pin RASEN_AGENT_RUNTIME per-test (dispatchMode assertion) - #3 bootstrap-obtain registry cleanup via updateStoreRegistryState (was regex corrupting the entry) - #6 board-page waitForState polling helper (fresh act() per poll) M1: lost-race diagnostic split (retry, not relocate). T2: theoretical TOCTOU accepted-known. Co-Authored-By: Claude Opus 5 (1M context) <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.
Why
Rasen's orchestration (subagent dispatch/resume, worker lifecycle) requires per-agent adaptation. Upstream OpenSpec shipped installers for ~27 code agents, but Rasen has only adapted Claude Code (first-class) and Codex — advertising the other installers is misleading. This PR narrows the install surface to adapted agents only, then adds Hermes as the third adapted agent.
What
agent-adapters-visible-tools (a2a52ce)
adapted?: booleanflag onAIToolOption; selection surface (getToolsWithSkillsDir()) narrows to adapted tools — interactive multi-select,--tools all, help text--tools <hidden>rejects with a distinct "recognized but not yet adapted" message (isKnownUnadaptedTool()); unknown tokens keep the old errorupdatenew-tool nudges filter to adapted; tolerant-read surfaces stay full-list so pre-existing installs (e.g..cursor/) keep refreshing viarasen updateagent-adapters-hermes (73b321e)
hermes-agent) as third adapted agent:adapted: true, skillsHome: 'global'src/core/hermes/hermes-home.ts(HERMES_HOMEenv or~/.hermes, mirrors codex-home)resolveToolSkillsRoot()threaded through 7 call sites (init/update/tool-detection/migration/profile-sync-drift) — project-local default byte-identical for existing tools; hermes resolves to~/.hermes/skillsquick_commandsare shell shortcuts, not prompts); this change's tests also close the previously-untested adapted-but-adapterless pathhermes -z,--resume) intentionally out of scope; documented for a future changeSpecs
+1 new capability (
hermes-integration), deltas synced intoadapted-agent-visibility(new),ai-tool-paths,cli-init,cli-update.rasen validate --specs --strict: 134 passed, 0 failed.Tests
Full suite 2720/2720 (was 2714; +6 hermes tests). Review: 2 rounds across children — round-1 Major (
CODEX_HOMEenv leak in tests) fixed and re-verified; accepted-known minors recorded in change archives.🤖 Generated with Claude Code