Feature/issue triage#8
Merged
Merged
Conversation
Create .github/workflows/agentry-planning-pipeline.yml triggered on issues:opened, running the planning-pipeline workflow with the github-actions binder. Update planning-pipeline.yaml with source/fallback inputs, issue:comment + issue:label capabilities, and an agent block. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ping, and pr:create Replace top-level model: block with agent: block (runtime: claude-code, max_iterations: 3). Add source: issue.body / fallback: issue.title to issue-description input so GitHub issue bodies are auto-resolved. Add pr:create and issue:comment to tools.capabilities. Expand bug-fix-system-prompt.md to instruct the agent to commit the fix with the originating issue number, open a PR with the agent-proposed label, and post a comment on the original issue linking to the fix PR. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Delete .github/workflows/agentry-issue-triage.yml which has been superseded by agentry-planning-pipeline.yml. Update the planning-pipeline workflow with a comment explaining that it replaces the triage-only workflow. Verification confirms no remaining references to the deleted workflow in .github/ or active workflow configurations. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Create .github/workflows/agentry-bug-fix.yml that triggers on issues labeled with 'category:bug' and runs the bug-fix workflow. Follows the established pattern from agentry-code-review.yml with proper permissions and secrets configuration. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add feature-implement.yaml with identity v1.0.0, claude-code agent (max_iterations: 10, timeout: 600s, elevated trust), issue.body source mapping with issue.title fallback, all six required tool capabilities (repository:read, shell:execute, pr:create, issue:comment, issue:label, issue:create), and an output schema with action enum (implemented|decomposed), pr_url, sub_issues, and reasoning fields. Add feature-implement-system-prompt.md instructing the agent to assess implementability via a <=5 files / <=500 lines heuristic, implement with tests and open a PR with agent-proposed label when in scope, or create scoped sub-issues with category:feature + agent-decomposed labels and comment on the parent issue when too large. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Create .github/workflows/agentry-feature-implement.yml that triggers on issues labeled with 'category:feature' and runs the feature-implement workflow. Follows the established pattern from agentry-bug-fix.yml with proper permissions and secrets configuration. Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
📝 WalkthroughWalkthroughAdds three GitHub Actions workflows (planning, bug-fix, feature-implement), converts several workflows to use an agent runtime (claude-code) with expanded tool capabilities (PR creation, issue comment/label), updates workflow triggers/permissions, enhances prompts for autonomous implementation, and removes many legacy proof/test artifacts and recordings. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant User as Issue Author
participant GH as GitHub (Issues & Actions)
participant Action as GitHub Actions Runner
participant Agentry as agentry Agent (claude-code)
participant Repo as Repository (git/PR)
User->>GH: create issue / add label (category:bug/category:feature)
GH->>Action: trigger workflow (planning / bug-fix / feature-implement)
Action->>Agentry: run agentry CLI (workflows/*.yaml) with inputs
Agentry->>Repo: read repository files (repository:read)
alt implementable
Agentry->>Repo: create commit + push
Agentry->>Repo: create PR (pr:create)
Agentry->>GH: post issue comment (issue:comment) linking PR
else decompose
Agentry->>GH: create sub-issues (issue:create) and label parent (issue:label)
Agentry->>GH: comment on parent issue
end
Agentry->>Action: return JSON result (action/pr_url/sub_issues/reasoning)
Action->>GH: surface results (comments/labels/PR)
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 inconclusive)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
Agentry Code ReviewTokens: 25,427 in / 2,090 out |
Agentry Code Review |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 6
🧹 Nitpick comments (5)
docs/specs/ci-workflows/04-proofs/T04-03-no-remaining-references.txt (1)
5-6: Make the proof command assertive instead of informational.The current command does not fail on unexpected matches, so it’s weak as a verification artifact.
Suggested change
-Command: grep -r agentry-issue-triage .github/ || echo "No references found" +Command: | + matches=$(grep -R "agentry-issue-triage" .github/ \ + | grep -v '^.github/workflows/agentry-planning-pipeline.yml:# This workflow replaces the superseded agentry-issue-triage.yml workflow\.$' || true) + test -z "$matches" && echo "No unexpected references found" || { echo "$matches"; exit 1; }🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@docs/specs/ci-workflows/04-proofs/T04-03-no-remaining-references.txt` around lines 5 - 6, Replace the informational grep command so it fails the CI on unexpected matches: instead of "grep -r agentry-issue-triage .github/ || echo 'No references found'", run a grep that excludes planning-pipeline.yml (or explicitly ignores its explanatory comment) and returns non‑zero on any match (e.g., use grep -r --exclude=planning-pipeline.yml agentry-issue-triage .github/ and if it finds matches print an error and exit 1); update the command string in the proof (the grep invocation referencing "agentry-issue-triage" and the mention of planning-pipeline.yml) so CI fails when stray references are present..github/workflows/agentry-planning-pipeline.yml (1)
27-29: Pin the Claude Code npm package version to stabilize CI.Installing
@anthropic-ai/claude-codewithout a version can cause unexpected workflow breakages after upstream releases. The latest stable version is 2.1.86.Suggested change
- - name: Install Claude Code - run: npm install -g `@anthropic-ai/claude-code` + - name: Install Claude Code + run: npm install -g `@anthropic-ai/claude-code`@2.1.86🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.github/workflows/agentry-planning-pipeline.yml around lines 27 - 29, Pin the global installation of the Claude Code package by updating the workflow step named "Install Claude Code" to install a fixed version of `@anthropic-ai/claude-code` (use 2.1.86); replace the current npm install -g `@anthropic-ai/claude-code` invocation with an explicit version specifier `@2.1.86` to stabilize CI..github/workflows/agentry-feature-implement.yml (1)
27-27: Add a job timeout to prevent stuck runner consumption.At Line 27, consider adding
timeout-minutesfor theagentryjob. This improves CI reliability and cost control when external calls hang.Proposed fix
agentry: if: github.event.label.name == 'category:feature' runs-on: ubuntu-latest + timeout-minutes: 30 steps:🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.github/workflows/agentry-feature-implement.yml at line 27, Add a job-level timeout to the agentry GitHub Actions job to prevent stuck runners: in the job block with id "agentry" (the block containing the runs-on: ubuntu-latest line), add a top-level key like "timeout-minutes: 30" (or an appropriate number for your workflow) so the job is automatically canceled after that duration..github/workflows/agentry-bug-fix.yml (1)
25-27: Add a job-level timeout to prevent hung runs.Line 25 through Line 27 define a long-running agent job with no
timeout-minutes. Add an explicit limit for reliability and cost control.Proposed workflow update
jobs: agentry: if: github.event.label.name == 'category:bug' runs-on: ubuntu-latest + timeout-minutes: 15🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.github/workflows/agentry-bug-fix.yml around lines 25 - 27, The "agentry" job currently has no timeout, which can lead to hung runs and extra cost; add a job-level timeout by inserting a "timeout-minutes: 30" (or other agreed limit) under the agentry job block (alongside "if:" and "runs-on:") so the workflow will be automatically terminated after the specified number of minutes.workflows/prompts/bug-fix-system-prompt.md (1)
25-26: Return PR metadata in output for machine traceability.Line 25 and Line 26 define output keys that omit
pr_url, even though the flow requires opening a PR. Adding PR metadata improves downstream automation and auditing.Proposed output contract tweak
-Output format: JSON object with keys `diagnosis`, `root_cause`, `suggested_fix`, and `confidence`. +Output format: JSON object with keys `diagnosis`, `root_cause`, `suggested_fix`, `confidence`, and `pr_url` (when a PR is opened).🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@workflows/prompts/bug-fix-system-prompt.md` around lines 25 - 26, Update the output contract string "Output format: JSON object with keys `diagnosis`, `root_cause`, `suggested_fix`, and `confidence`." to include a `pr_url` field, e.g. add `pr_url` to the top-level keys so downstream automation can trace the PR; keep the existing `suggested_fix` requirement (must contain `file`, `line`, and `change`) and ensure the prompt text clearly mandates `pr_url` be a fully qualified URL in the produced JSON.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@docs/specs/ci-workflows/03-proofs/T03-03-workflow-structure.txt`:
- Around line 18-20: The current boolean checks (steps_checkout, steps_python,
steps_agentry) use substring matching on str(step) which yields false positives;
update the checks to iterate the agentry job's steps and explicitly inspect
step.get('uses') and step.get('run') fields: for steps_checkout check if
step.get('uses', '').startswith('actions/checkout@'), for steps_python check if
step.get('uses', '').startswith('actions/setup-python@'), and for steps_agentry
check for a precise match in step.get('uses', '') or a recognizable command in
step.get('run', '') (e.g., the agentry action identifier or exact CLI
invocation); replace the any(...) expressions with these explicit field checks
so unrelated/malformed steps no longer pass.
In `@docs/specs/ci-workflows/03-proofs/T03-feature-proofs.md`:
- Around line 29-33: Add a language tag to the fenced code block containing the
agentry CLI command so the block is fenced as bash; locate the block showing
"agentry --output-format json run workflows/feature-implement.yaml ..." and
change the opening fence from "```" to "```bash" (the code block around that
command is what triggers MD040).
In `@docs/specs/ci-workflows/04-proofs/T04-proofs.md`:
- Line 30: The summary currently uses the shorthand "issues: [opened]" which is
not valid GitHub Actions trigger syntax; update the summary to the proper
workflow notation used in the file (e.g., use "on: issues: types: [opened]" or
the exact trigger form present in the workflow header) so the documentation
matches the actual workflow trigger (reference the trigger string "issues:
[opened]" and replace it with the correct "on: issues: types: [opened]" form).
In `@workflows/feature-implement.yaml`:
- Around line 41-59: The schema currently lists action, pr_url, and sub_issues
but does not enforce that pr_url is required when action is "implemented" or
that sub_issues is required when action is "decomposed"; update the schema by
adding a JSON Schema conditional (if/then/else) or oneOf block referencing the
existing properties (action, pr_url, sub_issues, reasoning) so that if:
{properties:{action:{const:"implemented"}}} then: {required:["pr_url"]} and if:
{properties:{action:{const:"decomposed"}}} then: {required:["sub_issues"]}; keep
the existing types and enums and ensure sub_issues remains an array of strings.
In `@workflows/prompts/bug-fix-system-prompt.md`:
- Around line 9-15: Insert a mandatory verification step between "Implement the
fix" and "Commit the fix" that requires running relevant automated tests and
checks (unit/integration tests, linters, CI script) and confirming they pass
before proceeding, and update the subsequent "Commit the fix" / "Open a pull
request" instructions to state commits/PRs must only be created after this
verification is successful; modify the items named "Implement the fix", "Commit
the fix", "Open a pull request", and "Post a comment" accordingly so the flow
enforces "run and pass relevant tests/checks first" before committing or opening
the PR.
In `@workflows/prompts/feature-implement-system-prompt.md`:
- Around line 8-9: Step 2 in
workflows/prompts/feature-implement-system-prompt.md incorrectly assumes
planning-pipeline comments will exist; update the prompt text to first check for
planning-pipeline issue comments and, if none are found, fall back to using the
pipeline file outputs described in workflows/planning-pipeline.yaml and
workflows/task-decompose.yaml (or to trigger/consume the task-decompose output)
as the implementation plan; mention both the comment check and the deterministic
fallback so the prompt does not depend solely on issue-comment publication.
---
Nitpick comments:
In @.github/workflows/agentry-bug-fix.yml:
- Around line 25-27: The "agentry" job currently has no timeout, which can lead
to hung runs and extra cost; add a job-level timeout by inserting a
"timeout-minutes: 30" (or other agreed limit) under the agentry job block
(alongside "if:" and "runs-on:") so the workflow will be automatically
terminated after the specified number of minutes.
In @.github/workflows/agentry-feature-implement.yml:
- Line 27: Add a job-level timeout to the agentry GitHub Actions job to prevent
stuck runners: in the job block with id "agentry" (the block containing the
runs-on: ubuntu-latest line), add a top-level key like "timeout-minutes: 30" (or
an appropriate number for your workflow) so the job is automatically canceled
after that duration.
In @.github/workflows/agentry-planning-pipeline.yml:
- Around line 27-29: Pin the global installation of the Claude Code package by
updating the workflow step named "Install Claude Code" to install a fixed
version of `@anthropic-ai/claude-code` (use 2.1.86); replace the current npm
install -g `@anthropic-ai/claude-code` invocation with an explicit version
specifier `@2.1.86` to stabilize CI.
In `@docs/specs/ci-workflows/04-proofs/T04-03-no-remaining-references.txt`:
- Around line 5-6: Replace the informational grep command so it fails the CI on
unexpected matches: instead of "grep -r agentry-issue-triage .github/ || echo
'No references found'", run a grep that excludes planning-pipeline.yml (or
explicitly ignores its explanatory comment) and returns non‑zero on any match
(e.g., use grep -r --exclude=planning-pipeline.yml agentry-issue-triage .github/
and if it finds matches print an error and exit 1); update the command string in
the proof (the grep invocation referencing "agentry-issue-triage" and the
mention of planning-pipeline.yml) so CI fails when stray references are present.
In `@workflows/prompts/bug-fix-system-prompt.md`:
- Around line 25-26: Update the output contract string "Output format: JSON
object with keys `diagnosis`, `root_cause`, `suggested_fix`, and `confidence`."
to include a `pr_url` field, e.g. add `pr_url` to the top-level keys so
downstream automation can trace the PR; keep the existing `suggested_fix`
requirement (must contain `file`, `line`, and `change`) and ensure the prompt
text clearly mandates `pr_url` be a fully qualified URL in the produced JSON.
🪄 Autofix (Beta)
✅ Autofix completed
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 06e663b7-e253-4fdd-8e1f-923c35c336c1
📒 Files selected for processing (31)
.github/workflows/agentry-bug-fix.yml.github/workflows/agentry-feature-implement.yml.github/workflows/agentry-planning-pipeline.ymldocs/specs/07-spec-issue-triage/02-proofs/T02-01-file.txtdocs/specs/07-spec-issue-triage/02-proofs/T02-02-file.txtdocs/specs/07-spec-issue-triage/02-proofs/T02-03-file.txtdocs/specs/07-spec-issue-triage/02-proofs/T02-04-file.txtdocs/specs/07-spec-issue-triage/02-proofs/T02-proofs.mddocs/specs/07-spec-issue-triggered-triage/02-proofs/T02-01-file.txtdocs/specs/07-spec-issue-triggered-triage/02-proofs/T02-02-file.txtdocs/specs/07-spec-issue-triggered-triage/02-proofs/T02-proofs.mddocs/specs/ci-workflows/03-proofs/T03-01-file-creation.txtdocs/specs/ci-workflows/03-proofs/T03-02-yaml-validation.txtdocs/specs/ci-workflows/03-proofs/T03-03-workflow-structure.txtdocs/specs/ci-workflows/03-proofs/T03-feature-01-file-creation.txtdocs/specs/ci-workflows/03-proofs/T03-feature-02-yaml-validation.txtdocs/specs/ci-workflows/03-proofs/T03-feature-03-workflow-structure.txtdocs/specs/ci-workflows/03-proofs/T03-feature-proofs.mddocs/specs/ci-workflows/03-proofs/T03-proofs.mddocs/specs/ci-workflows/04-proofs/T04-01-file-deletion.txtdocs/specs/ci-workflows/04-proofs/T04-02-workflow-updated.txtdocs/specs/ci-workflows/04-proofs/T04-03-no-remaining-references.txtdocs/specs/ci-workflows/04-proofs/T04-proofs.mddocs/specs/triage/01-proofs/T01-01-file.txtdocs/specs/triage/01-proofs/T01-02-file.txtdocs/specs/triage/01-proofs/T01-proofs.mdworkflows/bug-fix.yamlworkflows/feature-implement.yamlworkflows/planning-pipeline.yamlworkflows/prompts/bug-fix-system-prompt.mdworkflows/prompts/feature-implement-system-prompt.md
Comment on lines
+18
to
+20
| 'steps_checkout': any('checkout' in str(s) for s in workflow.get('jobs', {}).get('agentry', {}).get('steps', [])), | ||
| 'steps_python': any('setup-python' in str(s) for s in workflow.get('jobs', {}).get('agentry', {}).get('steps', [])), | ||
| 'steps_agentry': any('agentry' in str(s) for s in workflow.get('jobs', {}).get('agentry', {}).get('steps', [])), |
There was a problem hiding this comment.
Harden step detection to avoid false-positive PASS results.
At Line 18–Line 20, using substring checks on str(step) is too permissive. A malformed or unrelated step can still pass. Prefer explicit matching on uses/run fields.
Proposed fix
- 'steps_checkout': any('checkout' in str(s) for s in workflow.get('jobs', {}).get('agentry', {}).get('steps', [])),
- 'steps_python': any('setup-python' in str(s) for s in workflow.get('jobs', {}).get('agentry', {}).get('steps', [])),
- 'steps_agentry': any('agentry' in str(s) for s in workflow.get('jobs', {}).get('agentry', {}).get('steps', [])),
+ 'steps_checkout': any(
+ s.get('uses', '').startswith('actions/checkout@')
+ for s in workflow.get('jobs', {}).get('agentry', {}).get('steps', [])
+ if isinstance(s, dict)
+ ),
+ 'steps_python': any(
+ s.get('uses', '').startswith('actions/setup-python@')
+ for s in workflow.get('jobs', {}).get('agentry', {}).get('steps', [])
+ if isinstance(s, dict)
+ ),
+ 'steps_agentry': any(
+ 'agentry --output-format json run workflows/bug-fix.yaml' in s.get('run', '')
+ for s in workflow.get('jobs', {}).get('agentry', {}).get('steps', [])
+ if isinstance(s, dict)
+ ),🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@docs/specs/ci-workflows/03-proofs/T03-03-workflow-structure.txt` around lines
18 - 20, The current boolean checks (steps_checkout, steps_python,
steps_agentry) use substring matching on str(step) which yields false positives;
update the checks to iterate the agentry job's steps and explicitly inspect
step.get('uses') and step.get('run') fields: for steps_checkout check if
step.get('uses', '').startswith('actions/checkout@'), for steps_python check if
step.get('uses', '').startswith('actions/setup-python@'), and for steps_agentry
check for a precise match in step.get('uses', '') or a recognizable command in
step.get('run', '') (e.g., the agentry action identifier or exact CLI
invocation); replace the any(...) expressions with these explicit field checks
so unrelated/malformed steps no longer pass.
Comment on lines
+29
to
+33
| ``` | ||
| agentry --output-format json run workflows/feature-implement.yaml \ | ||
| --input repository-ref=. \ | ||
| --binder github-actions | ||
| ``` |
There was a problem hiding this comment.
Add a language tag to the fenced code block.
At Line 29, the code fence is missing a language identifier, which triggers MD040 and weakens markdown tooling support.
Proposed fix
-```
+```bash
agentry --output-format json run workflows/feature-implement.yaml \
--input repository-ref=. \
--binder github-actions</details>
<details>
<summary>🧰 Tools</summary>
<details>
<summary>🪛 markdownlint-cli2 (0.22.0)</summary>
[warning] 29-29: Fenced code blocks should have a language specified
(MD040, fenced-code-language)
</details>
</details>
<details>
<summary>🤖 Prompt for AI Agents</summary>
Verify each finding against the current code and only fix it if needed.
In @docs/specs/ci-workflows/03-proofs/T03-feature-proofs.md around lines 29 -
33, Add a language tag to the fenced code block containing the agentry CLI
command so the block is fenced as bash; locate the block showing "agentry
--output-format json run workflows/feature-implement.yaml ..." and change the
opening fence from "" to "bash" (the code block around that command is
what triggers MD040).
</details>
<!-- fingerprinting:phantom:triton:hawk:8549180a-b638-40f6-b477-2ce640dd19d7 -->
<!-- This is an auto-generated comment by CodeRabbit -->
| ## Implementation Details | ||
|
|
||
| The planning pipeline workflow (agentry-planning-pipeline.yml) is the replacement for the triage-only workflow. It: | ||
| - Triggers on the same event: `issues: [opened]` |
There was a problem hiding this comment.
Use accurate GitHub Actions trigger notation in the summary.
issues: [opened] is not the workflow syntax used in the actual file and may mislead readers.
Suggested change
-- Triggers on the same event: `issues: [opened]`
+- Triggers on the same event: `on.issues.types: [opened]`🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@docs/specs/ci-workflows/04-proofs/T04-proofs.md` at line 30, The summary
currently uses the shorthand "issues: [opened]" which is not valid GitHub
Actions trigger syntax; update the summary to the proper workflow notation used
in the file (e.g., use "on: issues: types: [opened]" or the exact trigger form
present in the workflow header) so the documentation matches the actual workflow
trigger (reference the trigger string "issues: [opened]" and replace it with the
correct "on: issues: types: [opened]" form).
Comment on lines
+8
to
+9
| 2. Check for any comments on the issue from the planning-pipeline, which may have already decomposed the feature into a task list. Use those tasks as your implementation plan if present. | ||
|
|
There was a problem hiding this comment.
Prompt depends on planning comments that may never exist.
Line 8 assumes planning-pipeline decomposition is available in issue comments, but current pipeline contracts (see workflows/planning-pipeline.yaml and workflows/task-decompose.yaml) describe file outputs, not guaranteed issue-comment publication. This can make Step 1 nondeterministic.
Proposed prompt fallback clarification
-2. Check for any comments on the issue from the planning-pipeline, which may have already decomposed the feature into a task list. Use those tasks as your implementation plan if present.
+2. Check for planning-pipeline decomposition in issue comments. If no decomposition comment exists, proceed from the issue body alone and explicitly note that absence in `reasoning`.📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| 2. Check for any comments on the issue from the planning-pipeline, which may have already decomposed the feature into a task list. Use those tasks as your implementation plan if present. | |
| 2. Check for planning-pipeline decomposition in issue comments. If no decomposition comment exists, proceed from the issue body alone and explicitly note that absence in `reasoning`. | |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@workflows/prompts/feature-implement-system-prompt.md` around lines 8 - 9,
Step 2 in workflows/prompts/feature-implement-system-prompt.md incorrectly
assumes planning-pipeline comments will exist; update the prompt text to first
check for planning-pipeline issue comments and, if none are found, fall back to
using the pipeline file outputs described in workflows/planning-pipeline.yaml
and workflows/task-decompose.yaml (or to trigger/consume the task-decompose
output) as the implementation plan; mention both the comment check and the
deterministic fallback so the prompt does not depend solely on issue-comment
publication.
- Harden feature-implement output schema with allOf conditional (pr_url required when implemented, sub_issues when decomposed) - Add validation step to bug-fix prompt before commit/PR creation - Feature-implement prompt: graceful fallback when no planning comments exist - Remove 41 proof artifact files from git tracking (already gitignored) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Agentry Code ReviewTokens: 31,236 in / 2,338 out |
The _format_output_comment method produced empty comments when the agent returned no structured output or a raw string response. Now: - Extracts JSON from markdown-fenced or preamble-wrapped text - Shows raw agent response when no structured JSON found - Displays diagnostic message when output is completely empty Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Agentry Code ReviewTokens: 32,482 in / 1,930 out |
There was a problem hiding this comment.
🧹 Nitpick comments (2)
src/agentry/binders/github_actions.py (2)
1029-1036: The fence regex could be slightly more permissive.The regex
r"```(?:json)?\s*\n(.*?)```"requires a newline after the opening fence. Some LLM outputs may omit the newline (e.g.,```json{"key": "value"}```). However, the subsequent brace-matching fallback (lines 1038-1049) handles this case, so this is a minor observation rather than an issue.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/agentry/binders/github_actions.py` around lines 1029 - 1036, Relax the fence regex used in the re.search call that assigns fence_match so it also matches cases where there is no newline after the opening ```json fence (currently it requires a newline). Update the pattern in the re.search (the expression that produces fence_match) to allow optional whitespace/newline after the opening fence and still capture the inner block (keep re.DOTALL and non-greedy capture), so `fence_match` will succeed for both ```json\n{...}``` and ```json{...}``` outputs; leave the existing brace-matching fallback untouched.
1007-1051: Consider movingimport reto module level.The
reimport at line 1018 is inside the method. While functionally correct, placing it at the module level (around line 10-15) would be more consistent with the codebase style and avoid re-importing on each call.♻️ Suggested change
At the top of the file (around line 14):
from pathlib import Path from typing import Any, cast +import re import httpxThen remove the import from inside the method:
def _extract_json_from_text(text: str) -> dict[str, Any] | None: ... - import re - # Try direct parse first.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/agentry/binders/github_actions.py` around lines 1007 - 1051, Move the inline import for the regex module out of the _extract_json_from_text method and place a single top-level "import re" with the other module imports; then remove the "import re" line from inside the _extract_json_from_text function so the function uses the module-level re symbol instead of re-importing on each call.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@src/agentry/binders/github_actions.py`:
- Around line 1029-1036: Relax the fence regex used in the re.search call that
assigns fence_match so it also matches cases where there is no newline after the
opening ```json fence (currently it requires a newline). Update the pattern in
the re.search (the expression that produces fence_match) to allow optional
whitespace/newline after the opening fence and still capture the inner block
(keep re.DOTALL and non-greedy capture), so `fence_match` will succeed for both
```json\n{...}``` and ```json{...}``` outputs; leave the existing brace-matching
fallback untouched.
- Around line 1007-1051: Move the inline import for the regex module out of the
_extract_json_from_text method and place a single top-level "import re" with the
other module imports; then remove the "import re" line from inside the
_extract_json_from_text function so the function uses the module-level re symbol
instead of re-importing on each call.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: c58862e9-6872-4d13-b4a7-40d0339801d1
📒 Files selected for processing (1)
src/agentry/binders/github_actions.py
|
Note Autofix is a beta feature. Expect some limitations and changes as we gather feedback and continue to improve it. Fixes Applied SuccessfullyFixed 1 file(s) based on 3 unresolved review comments. A stacked PR containing fixes has been created.
Time taken: |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary by CodeRabbit
New Features
Improvements
Bug Fixes