Repository navigation
Conversation
KaiEureka
force-pushed
the
codex/vm-execution-profiles
branch
from
October 24, 2026 10:16
505348e to
0bf3e80
Compare
KaiEureka
force-pushed
the
codex/fix-llm-affordance-discovery-20261024
branch
from
October 24, 2026 10:38
5a6fd06 to
3b6906d
Compare
KaiEureka
marked this pull request as ready for review
October 24, 2026 11:08
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Implemented capabilities were missing from model-visible context or contradicted by static instructions. For example, authored Workflows accepted an isolation name without listing configured providers, open Teams omitted their declared specialists, and serialized senders were told to call a submit tool they might not have.
This change exposes configured environment choices and read-only Workflow descriptions, makes skill catalogs and file-edit modes visible through their actual tool schemas, and aligns Team, Base, and Analyst guidance with the running configuration. Existing execution permissions, resource allocation checks, model settings, budgets, and solving algorithms are preserved.
The audit covers 19 core model-callable tools, four example-specific tools, role/profile composition, Workflow entry points, and the three provider tool conversions. Nine finding families and their reproductions are documented in
docs/llm-discoverability.mdand its Chinese translation. Runtime descriptions also identify tools whose existing sandbox requirement is unsatisfied by the active environment.All final-head GitHub checks pass, including Python 3.10–3.14, distribution artifacts, macOS integrity, browser/SQLite fixtures, security, and translation checks. Python 3.12 completed 10,108 passing tests with 74 skips. The 25 direct regressions pass; against the preceding dependency revision they produce 24 failures and one passing control. A source-archive wheel was independently installed, its changed package sources compared byte-for-byte with the reviewed tree, and SDK, skill, file-edit, and Workflow discovery probes passed.
The branch includes the updated VM entry-isolation inheritance fix. Its capability catalog and execution provider share the effective-default calculation. A parent example's translation link was also corrected to satisfy the existing English-public-text check.
Depends on #247 and targets
codex/vm-execution-profiles, so the VM-specific discovery changes build on that series without changing its branches.