chore: migrate remaining runner_env keys to env.runner

[ralphbean]

Summary

• harness/code.yaml, review.yaml, and retro.yaml were the last three harness files still using the legacy top-level runner_env key.
• fix.yaml, triage.yaml, prioritize.yaml, and scribe.yaml already use the env.runner / env.sandbox nesting, so this brings the rest in line.
• Vars nested under forge.github (tokens, REPO_FULL_NAME, issue/PR numbers, GITHUB_PR_URL, ORIGINATING_URL) stayed within the forge selector since they're GitHub-specific; code.yaml's generic CODE_ALLOWED_TARGET_BRANCHES moved to the root env.runner.
• Updated a few doc/comment references (docs/code.md, env/code.env, eval/scripts/run-fullsend.sh) that named the old key.

Test plan

• python3 -c "import yaml; yaml.safe_load(open(...))" on each edited harness YAML file
• grep -r runner_env returns no matches

[qodo-code-review]

PR Summary by Qodo

chore: migrate remaining runner_env to env.runner in harness YAMLs

⚙️ Configuration changes 📝 Documentation 🕐 10-20 Minutes

AI Description

• Migrate code/review/retro harness configs from legacy runner_env to env.runner.
• Preserve GitHub-specific variables under forge.github while aligning structure across harnesses.
• Update docs/comments and env templates to reference env.runner instead of runner_env.

Diagram

graph TD
  A["Harness YAMLs (code/review/retro)"] --> B(["Fullsend runner"]) --> C(["Sandbox"])
  A --> D["Forge GitHub config"] --> B
  B --> E(["Runner env (env.runner)"])
  F["eval/scripts/run-fullsend.sh"] --> B
  G["Docs & env templates"] --> A

Loading

High-Level Assessment

The direct migration to env.runner is the right approach: it standardizes harness configuration, matches existing harnesses already using env.runner/env.sandbox, and reduces cognitive overhead. A compatibility shim (supporting both runner_env and env.runner) could be considered if external consumers depended on runner_env, but this PR only updates in-repo harness files and references, so a clean cutover is preferable.

Files changed (6) +23 / -19

Documentation (3) +3 / -3

code.mdUpdate CODE_ALLOWED_TARGET_BRANCHES doc to reference env.runner +1/-1

Update CODE_ALLOWED_TARGET_BRANCHES doc to reference env.runner

• Adjusts documentation to say CODE_ALLOWED_TARGET_BRANCHES is set via env.runner in harness/code.yaml instead of runner_env.

docs/code.md

code.envFix comment referencing PUSH_TOKEN injection key +1/-1

Fix comment referencing PUSH_TOKEN injection key

• Updates a comment to refer to the write-enabled PUSH_TOKEN as coming from env.runner rather than runner_env, reinforcing that it never enters the sandbox.

env/code.env

run-fullsend.shUpdate eval script comment to env.runner terminology +1/-1

Update eval script comment to env.runner terminology

• Renames a comment describing harness environment variable references from runner_env to env.runner (no functional behavior change).

eval/scripts/run-fullsend.sh

Other (3) +20 / -16

code.yamlMigrate runner-only vars to env.runner for code harness +9/-7

Migrate runner-only vars to env.runner for code harness

• Moves CODE_ALLOWED_TARGET_BRANCHES from top-level runner_env into env.runner. Also migrates the forge.github runner_env block (PUSH_TOKEN, repo metadata, issue number) into forge.github.env.runner.

harness/code.yaml

retro.yamlMigrate GitHub runner_env to forge.github.env.runner +5/-4

Migrate GitHub runner_env to forge.github.env.runner

• Replaces forge.github.runner_env with forge.github.env.runner for ORIGINATING_URL, REPO_FULL_NAME, and GH_TOKEN.

harness/retro.yaml

review.yamlMigrate GitHub runner_env to forge.github.env.runner +6/-5

Migrate GitHub runner_env to forge.github.env.runner

• Replaces forge.github.runner_env with forge.github.env.runner for REVIEW_TOKEN, REPO_FULL_NAME, PR_NUMBER, and GITHUB_PR_URL.

harness/review.yaml

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 6:30 PM UTC · Completed 6:44 PM UTC
Commit: 02dc8bb · View workflow run →

[qodo-code-review]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (1) 📜 Skill insights (0)

Context used

✅ Compliance rules (platform): 55 rules

✅ Skills: 4 invoked

  code-review

  code-implementation

  pr-review

  docs-review

1. Protected paths modified without issue ✗ Dismissed 📜 Skill insight § Compliance

Description

This PR modifies protected governance/infrastructure paths (harness/ and eval/scripts/), which
must never be auto-approved and require explicit authorization (linked issue/ADR) for the change
set. The PR description explains intent, but does not include a linked authorization artifact.

Code

harness/code.yaml[R51-53]

+env:
+  runner:
+    CODE_ALLOWED_TARGET_BRANCHES: "${CODE_ALLOWED_TARGET_BRANCHES}"

Relevance

●● Moderate

Protected-path rigor exists (PR #303), and justification links were requested/partly accepted before
(PR #184), but not consistent.

PR-#303
PR-#184

ⓘ Recommendations generated based on similar findings in past PRs

Evidence

Compliance rule 1538392 requires raising a finding whenever protected paths are modified, and
requires explicit authorization (linked issue/ADR) for such changes. The diff shows edits in
harness/ YAML files and eval/scripts/run-fullsend.sh, both listed as protected paths in the
rule.

harness/code.yaml[49-66]
harness/review.yaml[44-53]
harness/retro.yaml[40-48]
eval/scripts/run-fullsend.sh[148-166]
Skill: pr-review

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The PR modifies protected paths (`harness/`, `eval/scripts/`) but lacks an explicit linked authorization artifact (issue/ADR). This violates the requirement that protected governance/infrastructure changes must not be auto-approved and must be explicitly authorized.

## Issue Context
The PR is a mechanical migration from `runner_env` to `env.runner`, but it still changes protected infrastructure/configuration surfaces.

## Fix Focus Areas
- harness/code.yaml[49-66]
- harness/review.yaml[44-53]
- harness/retro.yaml[40-48]
- eval/scripts/run-fullsend.sh[148-166]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

2. Comment-only change in run-fullsend.sh 📘 Rule violation ⚙ Maintainability

Description

A comment was modified without any functional code change on the same or immediately adjacent lines
in the corresponding diff hunk, violating the requirement to avoid standalone comment-only edits
unless part of a refactor of the surrounding block.

Code

eval/scripts/run-fullsend.sh[156]

+  # Code/fix harness env.runner refs — mint normally sets these; eval skips mint.

Relevance

● Weak

Repo often merges standalone comment/doc wording updates (e.g., PR #171), so this comment-only edit
likely tolerated.

PR-#171

ⓘ Recommendations generated based on similar findings in past PRs

Evidence

Rule 1540690 requires that comment edits be paired with functional code changes in the same
hunk/block unless the block is being structurally refactored. The cited diff context shows only the
comment text changing (e.g., updating runner_env to env.runner) while the surrounding
logic/exports remain unchanged, making this a comment-only edit under the rule.

Rule 1540690: Do not change comments on lines without functional code changes in a diff
eval/scripts/run-fullsend.sh[152-166]
env/code.env[4-10]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
A comment line was updated without any accompanying functional code change in the same or adjacent lines in the diff hunk, which is not allowed unless it is part of a structural refactor of the surrounding block.

## Issue Context
The change updates wording from `runner_env` -> `env.runner`, but the surrounding logic/code (including exported variables) is otherwise unchanged in this PR.

## Fix Focus Areas
- eval/scripts/run-fullsend.sh[152-166]
- env/code.env[4-10]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

_{To customize comments, go to the Qodo configuration screen, or learn more in the docs.}

[fullsend-ai-review]

Review

Findings

High

• [protected-path] harness/code.yaml, harness/retro.yaml — PR modifies files under the protected harness/ path without a linked issue. Human approval is always required for changes to governance and infrastructure files.
  Remediation: Link an issue authorizing the harness file changes, or obtain explicit human approval.

Low

• [breaking-schema-change] harness/code.yaml:51 — Migration from runner_env to env.runner changes the harness YAML schema contract. The downstream consumer (fullsend-ai/fullsend) already supports this schema since fix.yaml, triage.yaml, prioritize.yaml, scribe.yaml, and review.yaml all use env.runner. Two separate runner_env blocks are migrated: one at top level (line 51) and one under forge.github (line 61).

• [breaking-schema-change] harness/retro.yaml:44 — Migration from runner_env to env.runner changes the harness YAML schema contract under forge.github. The downstream consumer already supports this schema since five other harness files use env.runner.

---

Next steps:

• /fs-fix — agent addresses review findings automatically
• /fs-fix <your instruction> — agent fixes with your specific guidance
• Push commits directly — review re-runs automatically on push
• /fs-fix-stop — disable automatic fix runs for this PR

Previous run

Review

Findings

High

• [protected-path] harness/code.yaml, harness/retro.yaml, harness/review.yaml — PR modifies files under the protected harness/ path without a linked issue. Human approval is always required for changes to governance and infrastructure files. The PR lacks a linked issue to justify the protected-path modifications.
  Remediation: Link an issue authorizing the harness file changes, or obtain explicit human approval.

Low

• [missing-authorization] Non-trivial structural change (42 lines, YAML schema migration) lacks a linked issue. This is a mechanical migration completing a pattern already established in fix.yaml, triage.yaml, prioritize.yaml, and scribe.yaml.

• [breaking-schema-change] harness/code.yaml:51 — Migration from runner_env to env.runner changes the harness YAML schema contract. The downstream consumer (fullsend-ai/fullsend) likely already supports this schema since fix.yaml, triage.yaml, prioritize.yaml, and scribe.yaml all use env.runner.

---

Next steps:

• /fs-fix — agent addresses review findings automatically
• /fs-fix <your instruction> — agent fixes with your specific guidance
• Push commits directly — review re-runs automatically on push
• /fs-fix-stop — disable automatic fix runs for this PR

[fullsend-ai-coder]

🤖 Finished Fix · ✅ Success · Started 6:45 PM UTC · Completed 6:49 PM UTC
Commit: 02dc8bb · View workflow run →

[fullsend-ai-coder]

🔧 Fix agent — iteration 1 (bot-triggered)

All 3 review findings are governance/process concerns (missing linked issue for protected-path modifications) that require human action, not code fixes. The fix agent cannot create or link issues, nor grant human approval. Protected-path enforcement in post-review.sh already prevents auto-approval for these paths — a human reviewer must approve. No code changes are needed.

Disagreed (3):

1. [protected-path] harness/ files modified without linked issue: This is a governance/process concern requiring human action (linking an issue or obtaining explicit human approval), not a code defect the fix agent can address. Protected-path enforcement already lives in post-review.sh — the review agent cannot approve PRs that touch harness/ paths, so a human reviewer must approve. This mechanism is working as designed. The fix agent cannot create or link issues.
2. [missing-authorization] Structural change lacks linked issue: This finding is the same governance concern as the protected-path finding — the PR needs a linked authorization artifact. Creating or linking issues to PRs is a human action outside the fix agent's scope. The reviewer notes this is a mechanical migration completing a pattern already established in fix.yaml, triage.yaml, prioritize.yaml, and scribe.yaml.
3. [breaking-schema-change] runner_env to env.runner schema migration: The reviewer acknowledges downstream (fullsend-ai/fullsend) likely already supports the env.runner schema since fix.yaml, triage.yaml, prioritize.yaml, and scribe.yaml all use it. This is informational context about the change's impact, not a code defect requiring a fix. The migration is completing an established pattern, not introducing a novel schema.

Tests: passed

Next steps:

• /fs-review — request a re-review of the changes
• /fs-fix <your instruction> — run another fix pass with specific guidance
• Push commits directly — review re-runs automatically on push
  _{Updated by fullsend fix agent}

[waynesun09]

Clean — no new findings from review squad. Remaining threads are addressed (schema migration already supported downstream) or explicitly deferred by the author with reasoning (ADR 080/081 compliance follow-up). CI green.

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 7:14 PM UTC · Completed 7:30 PM UTC
Commit: 4f665cc · View workflow run →

[fullsend-ai-review]

See the review comment for full details.

[fullsend-ai-review]

[low] breaking-schema-change

Migration from runner_env to env.runner changes the harness YAML schema contract. The downstream consumer (fullsend-ai/fullsend) already supports this schema since fix.yaml, triage.yaml, prioritize.yaml, scribe.yaml, and review.yaml all use env.runner. Two separate runner_env blocks are migrated: one at top level (line 51) and one under forge.github (line 61).

[fullsend-ai-review]

[low] breaking-schema-change

Migration from runner_env to env.runner changes the harness YAML schema contract under forge.github. The downstream consumer already supports this schema since five other harness files use env.runner.

[fullsend-ai-coder]

🤖 Finished Fix · ✅ Success · Started 7:30 PM UTC · Completed 7:32 PM UTC
Commit: 4f665cc · View workflow run →

[fullsend-ai-coder]

🔧 Fix agent — iteration 1 (bot-triggered)

PR #631 is already merged. No fixes were applied because the PR is no longer open.

Disagreed (1):

1. PR already merged: PR chore: migrate remaining runner_env keys to env.runner #631 is in MERGED state. No fixes can be applied to a merged PR. The review feedback is moot.

Tests: passed

Next steps:

• /fs-review — request a re-review of the changes
• /fs-fix <your instruction> — run another fix pass with specific guidance
• Push commits directly — review re-runs automatically on push
  _{Updated by fullsend fix agent}

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 7:34 PM UTC · Completed 7:45 PM UTC
Commit: 4f665cc · View workflow run →

[fullsend-ai-retro]

Retro: PR #631 — chore: migrate remaining runner_env keys to env.runner

PR #631 was a straightforward 5-file chore PR by the repo maintainer (ralphbean) migrating three harness YAML files from the legacy runner_env key to env.runner and updating doc/comment references. 17 additions, 14 deletions.

Timeline

Time (UTC)	Event	Duration	Outcome
18:29–18:44	First review (run 30841493456)	15 min	1 HIGH protected-path, 2 LOW breaking-schema-change. Verdict: CHANGES_REQUESTED
18:45–18:49	First fix (run 30842666698)	4.5 min	Correctly disagreed with all 3 findings as governance/process concerns. No code changes.
19:11	Human approval (waynesun09)	—	"Clean — no new findings from review squad."
19:12	Auto-merge enabled	—	—
19:13	Merge commit created (main → branch)	—	No code changes, only merge reconciliation
19:14–19:30	Second review (run 30844865718)	16 min	Re-raised same breaking-schema-change findings. Verdict: CHANGES_REQUESTED
19:27	PR entered merge queue	—	—
19:30–19:32	Second fix (run 30846063558)	2.5 min	PR already merged; stopped immediately.
19:32	PR merged	—	—

Total wall-clock: ~63 minutes from PR open to merge.

What went well

• Fix agent performed correctly in both cycles — disagreed with governance findings on the first run, detected the merged state and stopped cleanly on the second.
• Human reviewer (waynesun09) provided an accurate, well-calibrated approval.
• The PR was correct as submitted and merged without any rework.

Waste identified

The second review + fix cycle was entirely unnecessary (~18.5 min of Opus compute). The merge commit (merging main into the branch) contained no code changes but triggered a synchronize event that dispatched a full review. That review re-raised the same breaking-schema-change findings and produced a CHANGES_REQUESTED verdict on a PR that was already human-approved and entering the merge queue. The subsequent fix agent found the PR already merged.

Evidence for existing issues (no new proposals needed)

• fullsend#1282 — Skip re-review on merge-only commits: the merge commit at 19:13 added zero code changes but triggered a 16-min Opus review. This is the third observed instance of this pattern.
• agents#106 — Review re-raises findings after author declines scope: the author addressed the breaking-schema-change findings in PR comments (19:21–19:24), but the second review re-raised them at 19:30.
• fullsend#2115 — COMMENT verdict for human-authored PRs with only medium/low findings: the first review used CHANGES_REQUESTED due to the HIGH protected-path finding, triggering the fix agent on what were purely governance concerns for a maintainer-authored chore PR.
• agents#301 — Lightweight review for trivial/mechanical changes: this 17-line config migration following an established pattern received two full 6-dimension Opus reviews.
• fullsend#3690 — Harness pre-script abort when PR is merged: the second fix agent launched despite the PR being in the merge queue, then found it merged and stopped after ~2.5 min.