- Home
- Skills
- Workflow Automation
- Codex Parallel Workflows
Codex Parallel Workflows
Run independent code reviews and document jobs in parallel with Codex CLI. Includes two English workflows and reusable task templates for the free codex-parallel runner on Linux.
$5
Works with the AI Agents you already use
Codex Parallel Workflows
Example session with this skill installed
Review the pricing and shipping modules independently for incorrect behavior in ordinary use. Keep source unchanged. Return each confirmed issue with its source location, a normal input, expected behavior, observed behavior, and evidence.
- Read your context and instructions
- Compiled the codex parallel workflows
See it in action
A real run on two deliberately small example modules, using the released runner and the module-review workflow. The source files contain ordinary pricing and shipping mistakes for demonstration; this is not a production audit or a benchmark.
Request
Review the pricing and shipping modules independently for incorrect behavior in ordinary use. Keep source unchanged. Return each confirmed issue with its source location, a normal input, expected behavior, observed behavior, and evidence.
The two source modules used in this run are shown below. Their documented behavior is the acceptance criterion for the review.
Pricing source
"""Order pricing: apply discount_percent to the entire subtotal."""
def order_total(unit_price, quantity, discount_percent=0):
"""Return the total after a percentage discount, rounded to two decimals."""
subtotal = unit_price * quantity
return round(subtotal - discount_percent / 100, 2)
Shipping source
"""Shipping costs 5.00, or is free for subtotals of at least 50.00."""
def shipping_fee(subtotal):
"""Return the shipping charge for an order subtotal."""
return 0.0 if subtotal > 50.0 else 5.0
Observed run
- Runner:
5f886cbbd9f9843a5793e2b060c4cf05e744a601. - Model:
gpt-6.1-sol; reasoning effort:low; concurrency limit:
- This is the validation account's model, not a model availability promise.
- Both tasks completed on their first attempt. Final
report.okwas true; attention count was 0. - The parent independently repeated the numerical checks and confirmed that the source files were unchanged.
Complete worker answers
The answer bodies below are complete. File-link targets have been changed to the source sections above; no findings or caveats have been removed.
review-pricing
One confirmed finding
Percentage discounts are calculated incorrectly — pricing.py:7. The documented contract applies the discount to the entire subtotal, but the implementation subtracts only discount_percent / 100.
- Normal input:
order_total(20, 3, 10). - Expected:
54.00(60 subtotal minus 10%). - Observed:
59.90. - Checked by executing the unchanged source with Python 3 standard-library facilities. A second case,
order_total(19.99, 2, 25), returned39.73instead of29.98. The no-discount control returned the expected60.00.
No pricing-specific tests were found in the supplied repository. Source inputs were unchanged; no unverified findings are reported.
review-shipping
One confirmed finding
Free-shipping threshold excludes exactly 50.00 — shipping.py:6. Normal input: shipping_fee(50.0). Expected: 0.0, because the documented contract promises free shipping for subtotals “of at least 50.00.” Observed: 5.0, because the comparison uses > instead of >=. Confirmed by executing the unchanged source with Python 3 standard-library facilities. Checks at 0.00, 49.99, 50.01, and 100.00 returned the expected charges.
No shipping-specific tests were found in the fixture or repository test search. Source inputs were unchanged; no reproduction files were created.
Combined result
The pricing module subtracts a small fixed amount instead of applying the percentage to the subtotal. The shipping module charges for an order of exactly 50.00 even though its documented threshold includes 50.00. Both assigned reviews completed. No source edits were requested or made.
Connects securely to your tools. The creator never sees your data.
About this skill
Give your coding agent a set of independent jobs and collect a result for each one. Codex Parallel Workflows guides the agent through defining task boundaries, submitting a batch with a concurrency limit, collecting answers as they arrive, and checking the final results.
The skill includes two English workflows:
- Review several modules. Assign one read-only review per module. Collect findings with file references, then combine them into a review that distinguishes confirmed issues from work still to check.
- Process a folder of documents. Apply the same instructions to each document, keep outputs separate, and check coverage before combining the results.
Each workflow shows how to structure the task manifest, choose a model and reasoning effort, and handle jobs that finish incomplete or need another attempt. The runner's optional terminal panel shows each task's status, prompt, and final answer.
Separate setup is required. These instructions use codex-parallel, a free MIT-licensed Linux runner. Its GitHub repository already includes a Codex skill that can run batches. This listing provides English workflows, examples, and templates; it does not include the runner, install it for you, or unlock additional runner features.
- Suited to jobs that can run independently. Tasks that depend on earlier output need a separate batch, and parallel workers should not edit the same files.
- The runner checks execution and result status. Your agent still checks the findings and generated files before treating the work as complete.
- The runner requires a Linux environment with local command execution. Loading the skill through MCP does not supply that environment, install the runner, or replace Codex CLI.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 10 days ago
Requires
Permissions
Allowed Hosts
Uses the buyer’s existing Codex authentication and model allowance. Selected prompts and task input may be sent to the Codex model service. Reads selected source files and writes manifests, logs and outputs in a dedicated batch workspace. The module-review workflow leaves source unchanged. The package contains instructions and templates only; it does not install the runner or include model credits.