Rectify: Authoritative Backend Selection and Portable Skill Semantics - #4454
Merged
Trecek merged 34 commits intoAug 4, 2026
Merged
Conversation
Trecek
force-pushed
the
impl-rectify-backend-authority-portable-semantics-20260801-134319
branch
from
August 4, 2026 16:19
4eb9456 to
067673a
Compare
Trecek
enabled auto-merge
August 4, 2026 20:18
Trecek
deleted the
impl-rectify-backend-authority-portable-semantics-20260801-134319
branch
August 4, 2026 20:36
Trecek
added a commit
that referenced
this pull request
Aug 4, 2026
…4462) #4454 retired the lexical capability vocabulary but migrated only the two .claude/skills/ entries with contract tests, stranding every other project-local skill: chart-course, check-bearing, review-promotion, audit-feature-gates, and eval-agent were rejected by the schema-v1 catalog at session start and dead as slash commands. Migrate all five to semantic_version 1: retired uses_capabilities tokens removed, semantic_requirements blocks added matching each skill's actual delegation structure (declared sibling_skills for cross-skill references), and backend-native body syntax (Agent(...), subagent_type=) rewritten to the portable phrasing #4454 established. test_eval_agent_uses_agent_tool asserted the retired Agent(subagent_type= string — a skill cannot both contain it and pass the schema-v1 catalog — so it is replaced by test_eval_agent_delegates_to_named_agent, asserting the declared evaluated-agent child_spawns role and the named-agent reference. Verified: all five resolve with invalid_reason=None under the installed 0.10.926 parser; cook tier gate passes; full-suite failure set is byte-identical to the pre-change baseline (61 pre-existing failures, unrelated — see #4462). Co-Authored-By: Claude Fable 5 <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.
Summary
Issue #4269 is a cross-layer authority failure, not a missing conditional. AutoSkillit currently lets lexical skill metadata, a model token interpreted as a provider profile, and raw provider environment keys replace the selected backend. Admission, dispatch, direct launchers, fleet quota, retry/resume, and evidence then reconstruct different fragments of the launch decision. Direct Codex cook exposes a second face of the same weakness: projection removes capability metadata but leaves Claude-shaped collaboration syntax in the skill body, so a capable Codex parent can refuse or silently degrade mandatory multi-agent work.
Implementation Plan
Plan file:
/home/talon/projects/generic_automation_mcp/.autoskillit/temp/rectify/rectify_backend_authority_and_portable_skill_semantics_2026-08-01_114622.md🤖 Generated with Claude Code via AutoSkillit