Skip to content

docs(#248): add review guidance for inconsistent backoff strategies - #260

Closed
fullsend-ai-coder[bot] wants to merge 2 commits into
mainfrom
agent/248-review-backoff-consistency
Closed

docs(#248): add review guidance for inconsistent backoff strategies#260
fullsend-ai-coder[bot] wants to merge 2 commits into
mainfrom
agent/248-review-backoff-consistency

Conversation

@fullsend-ai-coder

Copy link
Copy Markdown

Add a "Code pattern consistency" section to AGENTS.md instructing the review agent to flag inconsistent retry/backoff strategies within a single file as a low-severity informational finding. When a PR introduces or modifies a retry loop in a file that already has one, the reviewer should compare strategies (additive, exponential, constant, jitter-based) and ask the author to harmonize or justify the divergence.

Excludes from flagging: files with only one pattern, same-type strategies differing only in parameters, and cross-file divergence.

Note: make lint could not complete in the sandbox due to shellcheck-py installation failure (network 403). The change is documentation-only (AGENTS.md). Secret scan passed.


Closes #248

Post-script verification

  • Branch is not main/master (agent/248-review-backoff-consistency)
  • Secret scan passed (scan-secrets AGENTS.md)
  • Tests ran inside sandbox (N/A — docs-only change)

Add a "Code pattern consistency" section to AGENTS.md instructing the
review agent to flag inconsistent retry/backoff strategies within a
single file as a low-severity informational finding.
@fullsend-ai-review

fullsend-ai-review Bot commented Jul 10, 2026

Copy link
Copy Markdown

🤖 Finished Review · ❌ Failure · Started 12:50 PM UTC · Completed 12:51 PM UTC
Commit: c776c24 · View workflow run →

@guyoron1

Copy link
Copy Markdown
Owner

/fs-review

@fullsend-ai-review

fullsend-ai-review Bot commented Jul 10, 2026

Copy link
Copy Markdown

🤖 Finished Review · ❌ Failure · Started 3:57 PM UTC · Completed 4:08 PM UTC
Commit: d8e3df7 · View workflow run →

@guyoron1

Copy link
Copy Markdown
Owner

/fs-review

@fullsend-ai-review

fullsend-ai-review Bot commented Jul 10, 2026

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 4:31 PM UTC · Completed 4:42 PM UTC
Commit: d8e3df7 · View workflow run →

@fullsend-ai-review

fullsend-ai-review Bot commented Jul 10, 2026

Copy link
Copy Markdown

Review — Approve

PR #260 adds a "Code pattern consistency" section to AGENTS.md with review guidance for flagging inconsistent retry/backoff strategies within a single file. This is a docs-only change (+12 lines) that traces directly to issue #248.

Dimensions evaluated

Dimension Result
Correctness 1 low finding
Security No findings
Intent & coherence Authorized, scoped correctly
Style & conventions Consistent with existing patterns
Documentation currency No staleness
Cross-repo contracts N/A (no API surface)

Findings

[low · terminology] AGENTS.md — The term "additive backoff" is not standard. The established term for a backoff strategy where delay increases by a fixed increment per retry (delay += step, producing 3s, 6s, 9s, ...) is linear backoff. "Additive" in the retry/backoff domain typically describes how jitter is applied to a backoff curve (additive jitter vs. multiplicative jitter), not the backoff strategy itself. Consider replacing "additive" with "linear" in both the example parenthetical and the strategy enumeration to match the terminology reviewers will encounter in code and documentation.

Notes

  • The change follows the existing **When reviewing PRs:** directive pattern used by the Forge abstraction and ADR sections.
  • The exclusion criteria (three "Do not flag when" bullets) are logically sound: single-pattern files are excluded, same-type/different-parameter variations are excluded, and cross-file divergence is explicitly out of scope.
  • The severity calibration (low/informational) is appropriately lower than the medium+ used for architectural violations in surrounding sections.
  • The "or explain why they differ" branch correctly handles cases where different contexts within a file legitimately require different strategies.

Reviewed dimensions: correctness · security · intent-coherence · style-conventions · docs-currency


Protected paths detected — this PR modifies files under one or more
protected paths. The review agent cannot approve PRs that touch these paths.
A human reviewer must approve this PR.

Protected files in this PR:

  • AGENTS.md

Labels: PR is a documentation-only change to AGENTS.md

Previous run

Review

Findings

Medium

Low

  • [naming-consistency] AGENTS.md:65 — Section title "Code pattern consistency" is broader than its current content, which addresses only retry/backoff pattern consistency. The broader title is defensible if the section will accumulate additional code-pattern rules over time. Consider renaming to "Retry and backoff pattern consistency" or adding a clarifying sentence.

@fullsend-ai-review fullsend-ai-review Bot added the requires-manual-review Review requires human judgment label Jul 10, 2026
@guyoron1

Copy link
Copy Markdown
Owner

/fs-fix

@fullsend-ai-coder

fullsend-ai-coder Bot commented Jul 10, 2026

Copy link
Copy Markdown
Author

🤖 Finished Fix · ✅ Success · Started 5:42 PM UTC · Completed 5:50 PM UTC
Commit: d8e3df7 · View workflow run →

Add introductory sentence explaining the section collects pattern rules
for review, currently covering retry/backoff strategies, with room to
grow. Addresses naming-consistency review finding on PR #260.

Addresses review feedback on #260
@fullsend-ai-coder

Copy link
Copy Markdown
Author

🔧 Fix agent — iteration 1 (human-triggered)

Addressed 1 of 2 review findings. Added a clarifying intro sentence to the 'Code pattern consistency' section in AGENTS.md to explain the section's scope (currently retry/backoff, extensible to other patterns). Disagreed with the protected-path finding as it is procedural, not a code fix.

Fixed (1):

  1. naming-consistency: section title broader than content (AGENTS.md): Added introductory sentence clarifying the section collects code-pattern rules for review and currently covers retry/backoff strategies, with room for additional rules over time.

Disagreed (1):

  1. protected-path: AGENTS.md requires human approval: This is a procedural note, not a code defect. The fix agent cannot grant human approval — the CODEOWNERS enforcement and requires-manual-review label already ensure a human reviewer is required. No code change needed.

Tests: passed

Updated by fullsend fix agent

@fullsend-ai-review

fullsend-ai-review Bot commented Jul 10, 2026

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 5:53 PM UTC · Completed 5:59 PM UTC
Commit: d8e3df7 · View workflow run →

@fullsend-ai-review fullsend-ai-review Bot added requires-manual-review Review requires human judgment documentation Improvements or additions to documentation and removed requires-manual-review Review requires human judgment labels Jul 10, 2026
@guyoron1 guyoron1 closed this Jul 12, 2026
@guyoron1
guyoron1 deleted the agent/248-review-backoff-consistency branch July 12, 2026 03:10
guyoron1 pushed a commit that referenced this pull request Sep 10, 2026
Add an OpenHands section to docs/landscape.md in the production
agent orchestration systems category. Documents the platform's
architecture, the PolyForm licensing constraint on the enterprise
directory (only the MIT-licensed resolver and CLI are viable for
open-source use), known prompt injection vulnerabilities disclosed
by Johann Rehberger in 2025, and relevance to fullsend's problem
areas including sandboxing, injection defense, and observability.

Cross-references issue #260 which tracks concrete experiments
against the resolver.

Closes #55
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

documentation Improvements or additions to documentation requires-manual-review Review requires human judgment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[RTK+Ponytail] Review agent: flag inconsistent retry/backoff strategies within a file

1 participant