refactor: overhaul skill discovery and authorized workflows - #90
refactor: overhaul skill discovery and authorized workflows#90parsakhaz wants to merge 2 commits into
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![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: ece36a64c5
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".
| Use the repository's documented validation commands for the affected behavior. | ||
| Inspect current-head evidence already supplied by the implementation owner; | ||
| rerun when independent proof is required, inputs changed, or the result does | ||
| not resolve a review concern. Do not substitute hardcoded npm commands, require | ||
| irrelevant builds, or rerun a clean suite solely because a new review began. | ||
| Report the command, result, and evidence source; distinguish unavailable checks |
There was a problem hiding this comment.
When this reviewer runs in a repository whose gates are not TypeScript lint/typecheck, the new repository-derived policy conflicts with the unchanged output and rules below: lines 82-84 still require typecheck and lint, and line 149 explicitly requires running both. The Codex mirror retains the same conflict at lines 81-83 and 123, so either reviewer can invent irrelevant commands or misreport the checks that actually ran. Update the downstream report fields and rules to enumerate the discovered checks.
AGENTS.md reference: AGENTS.md:L27-L28
Useful? React with 👍 / 👎.
| Discover the repository's validation commands and inspect available evidence | ||
| for the reviewed head. Run checks that resolve a concrete correctness concern | ||
| or required gate; rerun after relevant changes. Reuse current passing results | ||
| otherwise. Record failures and unavailable checks without assuming a particular | ||
| package manager, application layout, or language. |
There was a problem hiding this comment.
For a non-Node review target, this now correctly discovers repository-specific validation such as go test or pytest, but the required review body at lines 104-106 still exposes only Typecheck and Lint. The posted review will therefore omit the checks that ran or present irrelevant fixed fields as its quality-gate evidence; make those rows dynamic or otherwise align the output contract with this generic gate policy.
AGENTS.md reference: AGENTS.md:L27-L28
Useful? React with 👍 / 👎.
|
|
||
| ## Render & Validate (MANDATORY) | ||
|
|
||
| You cannot judge a diagram from JSON alone. After generating or editing the Excalidraw JSON, you MUST render it to PNG, view the image, and fix what you see — in a loop until it's right. This is a core part of the workflow, not a final check. |
There was a problem hiding this comment.
This newly added Codex instruction uses an em dash despite the repository-wide authoring rule, and the same character recurs throughout the added Codex diagram, QA, and RunPane reference files. These files are installed as active instructions, so replace the em dashes in the new instructional prose rather than shipping guidance that violates its own authoring contract.
AGENTS.md reference: AGENTS.md:L29-L29
Useful? React with 👍 / 👎.
1 participant
The active workflows accumulated broad skill triggers, large always-loaded instructions, repeated validation recipes, and handoff gates that could stop an already-authorized implementation before its PR. This applies the principles in Eric Provencher's article across the active Parsa skill set.
What changes
The frozen tyler ancestor is untouched. Existing model/tool metadata and explicit invocation policies are retained, except simple-plan's read-only tool shortcut, which did not cover its implementation stage. Merge, release, deployment, production mutation, and release-asset publishing still need their matching authorization.
Documentation and diagrams
Rendering used Excalidraw 0.18's SVG exporter with bundled fonts and native SVG rasterization because local Chromium could not launch. No renderer or dependency changes are included.
Validation
This validates representative behavior and installation contracts. It does not claim a live multi-issue Pane run or production QA.