docs(changelog): soften resolver + smart-binding prose per security-disclosure standard - #175
Open
Coding-Dev-Tools wants to merge 2 commits into
Open
Conversation
…isclosure standard The 3rd-pass review's A5 audit flagged five phrases in this PR's CHANGELOG [Unreleased] entry that, taken together, broadcast a now-fixed weakness with a reproducible fingerprint and read as a regression target for anyone who can replay the eval pipeline. The user memory mem_01M0WKX6TR2YVRW12BRSAPHP4S paraphrases the pre-flagged "Do not broadcast security improvements or vulnerable details in changelogs" rule. The headline numbers the user explicitly wants preserved (k=50, savings_ratio 0.0 -> 0.4975, 35/36, 0/36, MRR 0.944 / 0.981, ~15 ms, 5/5 vs 0/5) are unchanged. Concrete redactions, by line in the original PR1 CHANGELOG: - Line 17 (Added): drops "Without this, every keyed write silently stored empty keys because the served gateway surface dropped the parameters." Re-narrates the Fixed bug; capability sentence kept. - Line 39 (Changed): drops the "(was 1/5 on the unkeyed benchmark pairs)" parenthetical. 35/36 and 0/36 numbers stay; the before/after ratio is what made the entry a regression-target tell. - Line 35 (Changed): reframes "without a stable `subject_key`" as a positive capability description (the aligned token-diff contract with explicit vetoes). The user pre-flagged this phrasing as a guide to a (now-fixed) weakness. - Line 53 (Fixed): drops "this is the underlying cause of the benchmark correction-miss pattern that the reworded-correction detector then had to compensate for." The two-flag coupling with line 35 reconstructed the silent-drop story end to end. Also adds one line to the cross-encoder bullet naming the file (`.env`) and the restart step explicitly, so the operator-facing enablement is unambiguous (per A2's readiness review). Owner-authored sections of CHANGELOG.md (Galaxy/dashboard/graph changelog entries below mine) are unchanged. CHANGELOG.md only. +14 -18. Follow-up to PR #171. Targeting ship/pr1-core-and-integration.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2d9a8012dd
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| regressions, MRR@5 lifts 0.889 -> 0.944 (sample) and 0.962 -> 0.981 (codemem), | ||
| with ~15 ms per query added. Not the default; flip with a one-line config. | ||
| with ~15 ms per query added. Not the default; flip with a one-line config | ||
| (`ENGRAPHIS_RERANK_MODEL` in `.env`, then restart the MCP server and dashboard). |
There was a problem hiding this comment.
When an operator follows this new instruction in a normal dashboard/MCP installation, Engraphis deliberately does not search the working directory for .env; it loads process variables or the owner-private ~/.engraphis/config.env instead. Restarting after editing ./.env therefore leaves ENGRAPHIS_RERANK_MODEL unset and the identity reranker active, so direct users to the trusted config file or process environment.
AGENTS.md reference: AGENTS.md:L40-L40
Useful? React with 👍 / 👎.
Per the open P2 review on PR 175: Engraphis deliberately does not
search the CWD for `.env` — process variables and the owner-private
`~/.engraphis/config.env` are the trusted sources. The previous
sentence ("ENGRAPHIS_RERANK_MODEL in `.env`, then restart ...")
would direct an operator to edit `./.env` and leave the identity
reranker active after restart. Now reads:
"set the value in the trusted config file
(`~/.engraphis/config.env` on the operator account, or as a
process environment variable) — Engraphis deliberately does
not read the CWD `.env`, so editing `./.env` and restarting
leaves the identity reranker active. Restart the MCP server
and dashboard after the change."
Co-authored-by: CommandCodeBot <noreply@commandcode.ai>
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.
Follow-up to #171. 3rd-pass review (8-agent audit) flagged five phrases in the
PR1 [Unreleased] CHANGELOG entry that, taken together, broadcast a now-fixed
weakness with a reproducible fingerprint and read as a regression target for
anyone who can replay the eval pipeline.
The user memory paraphrases the pre-flagged "Do not broadcast security
improvements or vulnerable details in changelogs" rule. Headline numbers the
user wants preserved (k=50, savings_ratio 0.0 -> 0.4975, 35/36, 0/36, MRR 0.944 /
0.981, ~15 ms, 5/5 vs 0/5) are unchanged.
Concrete redactions, by line in the original PR1 CHANGELOG:
empty keys" re-narration; capability sentence kept.
parenthetical. 35/36 and 0/36 numbers stay.
subject_key" as a positivecapability description (aligned token-diff contract with explicit vetoes).
correction-miss pattern that the reworded-correction detector then had to
compensate for."
Also adds one line to the cross-encoder bullet naming
.envand the restartstep explicitly, so the operator-facing enablement is unambiguous (per A2's
readiness review).
CHANGELOG.md only. +14 -18. Targeting
ship/pr1-core-and-integration(thePR1 branch, current tip 6cdfc57). Owner-authored sections of CHANGELOG.md
(Galaxy/dashboard/graph entries) are unchanged.