Repository navigation
Conversation
…tion-substrate gaps
The evidence-table commands for the /bf: and Skill("bf:...") counts
started self-matching once this document quoted those patterns in its
own prose, inflating the live grep results (14/4/7) above the recorded
values (13/3/5). Add --exclude=codex-compatibility.md so the commands
are reproducible against the current tree.
…c and probe 3 - Scope the $ARGUMENTS footnote's grep to skills/ (repo-wide count was inflated by this build's own files discussing the token) and add the scoped-all-file-types row to the canonical evidence table. - Rework probe 3's substitution instruction so the sentinel is only emittable from what the agent actually sees in the file content at read time, not from prose knowledge of the invocation text - consistent with how probe 2 handles the same class of risk. - Correct two [documented] grade tags to [inferred] where the claim is a reasonable conclusion rather than something Codex's docs state, and adjust the surrounding section language to match. - Replace internal FR6b/NFR1 spec-ID references in the doc's prose and headings with plain descriptive language. - Reword the doc's orchestration-substrate checklist entry to avoid "confirms" describing Codex-side runtime capability, while leaving the documented-absence phrasing on the adjacent line as-is (an explicitly permitted subject).
…codex doc - probe 3: drop the pre-printed "hello world" example from the probe body and checklist; require a run-time nonce chosen by the tester instead, so a passing transcript actually depends on Codex performing $ARGUMENTS substitution rather than being derivable from the probe file alone. - add a missing [inferred] example to the "Known grading" roster. - relabel the Provenance/evidence-table commit as the branch-base (merge-base with main) commit it actually is, rather than implying it tracks current HEAD.
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.
Answers "is it possible to make bf work with Codex too?" honestly rather than assuming — adds a repo-grounded portability document (docs/codex-compatibility.md) inventorying every place bf is coupled to Claude Code's host (plugin-root variable, cross-skill invocation, nested skill files, non-standard frontmatter), graded by confidence (documented/inferred/unconfirmed) since nobody in this session had a Codex CLI install to verify against. Ships alongside five small, explicitly-labeled untested candidate probe files (codex-probes/) that isolate the concrete mechanical unknowns, plus a 6-entry verification checklist so whoever has Codex CLI can resolve them in minutes. No existing skill file or plugin manifest was touched — this is investigation, not a port.