Add Copilot cascade routing: multi-agent PR review at $0 by Copilot · Pull Request #2 · BlackRoad-OS/simulation-hypothesis

Copilot · 2026-02-25T19:15:00Z

No .github/ configuration existed, leaving PRs without automated review and no Copilot agent guidance for the repo.

Changes

.github/copilot-instructions.md — Repository-wide context for Copilot agents: structure, languages, runtime, and validation steps. Reduces redundant exploration on every session.
.github/workflows/cascade-review.yml — Triggers on non-draft PR open/sync/reopen. Runs the cascade script then posts the combined review as a PR comment using gh.
.github/scripts/cascade_review.py — Core routing logic. Fetches the PR diff, fans out to two independent free-tier GitHub Models agents with distinct review personas, and writes a combined Markdown report to /tmp/cascade_review.md.

Models used (both free-tier, $0)

Agent	Model	Focus
1	`openai/gpt-4o-mini`	Correctness & clarity
2	`meta/meta-llama-3.1-8b-instruct`	Logic, edge cases & docs

Both are called via https://models.inference.ai.azure.com/chat/completions authenticated with the workflow's GITHUB_TOKEN — no extra secrets or billing opt-in required.

Original prompt

Please fix the pull requests and copilot cascade to route to multiple agents for $0

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

Co-authored-by: blackboxprogramming <118287761+blackboxprogramming@users.noreply.github.com>

Copilot

Pull request overview

Adds repository automation for PR review by routing PR diffs through multiple free-tier GitHub Models “agents” and posting a combined review comment, plus repo-wide Copilot guidance to reduce repeated context discovery.

Changes:

Introduces a GitHub Actions workflow to trigger on non-draft PR events and post an automated review comment.
Adds a Python cascade router script that fetches PR diffs, calls two free-tier models, and produces a combined Markdown report.
Adds repository-level Copilot instructions describing repo structure, runtime, and validation steps.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 7 comments.

File	Description
`.github/workflows/cascade-review.yml`	New PR-triggered workflow to run cascade review and post results via `gh`.
`.github/scripts/cascade_review.py`	New routing script: fetch diff, call two models, write combined Markdown output.
`.github/copilot-instructions.md`	New agent guidance about repo structure/runtimes/validation (needs accuracy updates).

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-02-25T19:27:22Z

.github/workflows/cascade-review.yml

+on:
+  pull_request:
+    types: [opened, synchronize, reopened]
+
+permissions:
+  contents: read
+  pull-requests: write
+


This workflow is triggered by pull_request, but for PRs from forks the GITHUB_TOKEN will be read-only regardless of pull-requests: write, so the gh pr comment step will fail. Consider either skipping the comment/posting steps for forked PRs, or switching to pull_request_target with a safe checkout strategy (use base repo code only) if you need to comment on fork PRs.

Copilot · 2026-02-25T19:27:22Z

.github/workflows/cascade-review.yml

+        run: |
+          gh pr comment "$PR_NUMBER" --body-file /tmp/cascade_review.md


On synchronize this will create a new PR comment every push (gh pr comment always posts a new comment), which can quickly spam long-lived PRs. Prefer updating an existing “Cascade Review” comment (e.g., find by marker and edit) or using an action that supports create-or-update semantics.

Suggested change

run: |

gh pr comment "$PR_NUMBER" --body-file /tmp/cascade_review.md

REPO: ${{ github.repository }}

run: |

set -euo pipefail

MARKER=""

BODY_FILE="/tmp/cascade_review_with_marker.md"

# Prepend a stable marker so we can find/update this comment later.

{

echo "$MARKER"

cat /tmp/cascade_review.md

} > "$BODY_FILE"

# Look for an existing Cascade Review comment on this PR.

EXISTING_COMMENT_ID="$(

gh api \

"repos/$REPO/issues/$PR_NUMBER/comments" \

--paginate \

--jq "[.[] | select(.body | contains(\"$MARKER\"))][0].id // empty"

)"

if [ -n "${EXISTING_COMMENT_ID:-}" ]; then

# Update the existing comment.

gh api \

"repos/$REPO/issues/comments/$EXISTING_COMMENT_ID" \

-X PATCH \

-f body=@"$BODY_FILE"

else

# Create a new comment (first run on this PR).

gh api \

"repos/$REPO/issues/$PR_NUMBER/comments" \

-f body=@"$BODY_FILE"

fi

Copilot · 2026-02-25T19:27:23Z

.github/scripts/cascade_review.py

+        )
+        diff = result.stdout or ""
+        return diff[:MAX_DIFF_CHARS]
+    except Exception as exc:
+        return f"(diff unavailable: {exc})"


get_pr_diff ignores gh pr diff failures (non-zero returncode) and will silently return an empty/truncated diff, producing low-signal reviews. Check result.returncode (or run with check=True) and include stderr in the error so the workflow fails loudly when diff retrieval breaks.

Suggested change

)

diff = result.stdout or ""

return diff[:MAX_DIFF_CHARS]

except Exception as exc:

return f"(diff unavailable: {exc})"

check=True,

)

diff = result.stdout or ""

return diff[:MAX_DIFF_CHARS]

except subprocess.CalledProcessError as exc:

stderr = (exc.stderr or "").strip()

message = f"gh pr diff failed with exit code {exc.returncode}"

if stderr:

message += f": {stderr}"

raise RuntimeError(message) from exc

except Exception as exc:

raise RuntimeError(f"Unexpected error while fetching PR diff: {exc}") from exc

Copilot · 2026-02-25T19:27:23Z

.github/scripts/cascade_review.py

+def main() -> None:
+    token = os.environ.get("GITHUB_TOKEN") or os.environ.get("GH_TOKEN", "")
+    pr_number = os.environ.get("PR_NUMBER", "")
+    pr_title = os.environ.get("PR_TITLE", "(no title)")
+    pr_body = os.environ.get("PR_BODY", "")[:1000]
+    repo = os.environ.get("REPO", "")
+
+    diff = get_pr_diff(pr_number, repo)
+


main() proceeds even if PR_NUMBER, REPO, or the token are missing/empty, which will generate misleading “model unavailable” output and still post a comment. Validate these env vars up front and exit non-zero (so the workflow fails) when required inputs are not present.

Copilot · 2026-02-25T19:27:23Z

.github/scripts/cascade_review.py

+    user_prompt = (
+        f"**PR title:** {pr_title}\n\n"
+        f"**PR description:**\n{pr_body}\n\n"
+        f"**Diff (may be truncated):**\n```diff\n{diff}\n```\n\n"
+        "Please provide 3–5 bullet points of review feedback."
+    )


The generated Markdown wraps the PR diff in a fenced code block, but neither the PR body nor the diff are escaped for occurrences of triple backticks. If either contains ``` the resulting PR comment can render incorrectly. Consider escaping/backtick-neutralizing fence sequences before embedding.

Copilot · 2026-02-25T19:27:23Z

.github/copilot-instructions.md

+- `README.md` — Overview, abstract, and evidence index
+- `code/` — Python 3 demonstration scripts (no external dependencies except `matplotlib` for `lorenz.py`)
+- `evidence/` — Evidence index with references


This claims the code/ scripts have no external dependencies except matplotlib for lorenz.py, but code/hue_man.py imports numpy and scipy. Update the repo instructions so agents don’t assume a dependency-free runtime.

Copilot · 2026-02-25T19:27:24Z

.github/copilot-instructions.md

+
+## Validation
+
+There are no automated tests. To validate changes to Python scripts, run them directly and confirm they produce expected output without errors. There are no linting configs or CI pipelines defined in this repository.


This says there are “no linting configs or CI pipelines defined”, but this PR adds a GitHub Actions workflow under .github/workflows/. Update this section to avoid contradicting the repo’s actual CI setup.

Suggested change

There are no automated tests. To validate changes to Python scripts, run them directly and confirm they produce expected output without errors. There are no linting configs or CI pipelines defined in this repository.

There are no automated tests or dedicated linting configuration files. To validate changes to Python scripts, run them directly and confirm they produce expected output without errors. A basic GitHub Actions workflow exists under `.github/workflows/`, but it does not replace manual validation.

Initial plan

e879cde

Copilot AI assigned Copilot and blackboxprogramming Feb 25, 2026

Copilot started work on behalf of blackboxprogramming February 25, 2026 19:15 View session

feat: add Copilot cascade routing to multiple free agents at $0 cost

faa7b25

Co-authored-by: blackboxprogramming <118287761+blackboxprogramming@users.noreply.github.com>

Copilot AI changed the title ~~[WIP] Fix pull requests and copilot cascade routing to multiple agents~~ Add Copilot cascade routing: multi-agent PR review at $0 Feb 25, 2026

Copilot finished work on behalf of blackboxprogramming February 25, 2026 19:21

blackboxprogramming requested a review from Copilot February 25, 2026 19:21

Copilot started reviewing on behalf of blackboxprogramming February 25, 2026 19:22 View session

Copilot AI reviewed Feb 25, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add Copilot cascade routing: multi-agent PR review at $0#2

Add Copilot cascade routing: multi-agent PR review at $0#2
Copilot wants to merge 2 commits intomainfrom
copilot/fix-cascade-routing-issue

Copilot AI commented Feb 25, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		run: \|
		gh pr comment "$PR_NUMBER" --body-file /tmp/cascade_review.md

-        run: |
-          gh pr comment "$PR_NUMBER" --body-file /tmp/cascade_review.md
+          REPO: ${{ github.repository }}
+        run: |
+          set -euo pipefail
+          MARKER="<!-- CASCADE_REVIEW_COMMENT -->"
+          BODY_FILE="/tmp/cascade_review_with_marker.md"
+          # Prepend a stable marker so we can find/update this comment later.
+          {
+            echo "$MARKER"
+            cat /tmp/cascade_review.md
+          } > "$BODY_FILE"
+          # Look for an existing Cascade Review comment on this PR.
+          EXISTING_COMMENT_ID="$(
+            gh api \
+              "repos/$REPO/issues/$PR_NUMBER/comments" \
+              --paginate \
+              --jq "[.[] | select(.body | contains(\"$MARKER\"))][0].id // empty"
+          )"
+          if [ -n "${EXISTING_COMMENT_ID:-}" ]; then
+            # Update the existing comment.
+            gh api \
+              "repos/$REPO/issues/comments/$EXISTING_COMMENT_ID" \
+              -X PATCH \
+              -f body=@"$BODY_FILE"
+          else
+            # Create a new comment (first run on this PR).
+            gh api \
+              "repos/$REPO/issues/$PR_NUMBER/comments" \
+              -f body=@"$BODY_FILE"
+          fi

-        )
-        diff = result.stdout or ""
-        return diff[:MAX_DIFF_CHARS]
-    except Exception as exc:
-        return f"(diff unavailable: {exc})"
+            check=True,
+        )
+        diff = result.stdout or ""
+        return diff[:MAX_DIFF_CHARS]
+    except subprocess.CalledProcessError as exc:
+        stderr = (exc.stderr or "").strip()
+        message = f"gh pr diff failed with exit code {exc.returncode}"
+        if stderr:
+            message += f": {stderr}"
+        raise RuntimeError(message) from exc
+    except Exception as exc:
+        raise RuntimeError(f"Unexpected error while fetching PR diff: {exc}") from exc


		## Validation

		There are no automated tests. To validate changes to Python scripts, run them directly and confirm they produce expected output without errors. There are no linting configs or CI pipelines defined in this repository.

	There are no automated tests. To validate changes to Python scripts, run them directly and confirm they produce expected output without errors. There are no linting configs or CI pipelines defined in this repository.
	There are no automated tests or dedicated linting configuration files. To validate changes to Python scripts, run them directly and confirm they produce expected output without errors. A basic GitHub Actions workflow exists under `.github/workflows/`, but it does not replace manual validation.

Uh oh!

Conversation

Copilot AI commented Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

Models used (both free-tier, $0)

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Copilot AI commented Feb 25, 2026 •

edited

Loading