t1313: Migrate dispatch.sh decision logic to AI#2224
Conversation
Replace three deterministic decision functions with AI-first equivalents: - classify_task_complexity(): 180-line keyword pattern matching → AI classification (sends description + tags to sonnet, gets tier back) - should_prompt_repeat(): heuristic retry eligibility → AI retry advisor (gathers failure reason, attempt history, pattern data → AI decides) - check_output_quality(): 6 separate heuristic checks → AI quality assessment (gathers log signals, diff stats, error patterns → AI assesses) Simplify resolve_task_model(): - Remove 5-step heuristic chain (pattern-tracker, keyword classification, cost-efficiency check, contest detection, budget degradation) - Replace with: AI classification + budget cap post-processing - Keep mechanical lookups (explicit model, subagent frontmatter, contest mode) All AI functions follow the proven gather→decide→execute pattern from ai-lifecycle.sh. Falls back to 'opus' if AI is unavailable. Net: -98 lines (3776 → 3678)
WalkthroughThe dispatcher now employs AI-first classification to determine task complexity and assign appropriate models (haiku/sonnet/opus), replacing pattern-based heuristics. AI evaluates task descriptions and outputs JSON-formatted tier assignments with fallback logic when unavailable; quality gates similarly delegate pass/fail decisions to AI rather than fixed pattern checks. Changes
Sequence DiagramsequenceDiagram
participant Task as Task Input
participant Dispatch as Dispatch Logic
participant AI as AI Classifier
participant ModelSel as Model Selection
participant Executor as Task Executor
participant QualityAI as Quality Gate AI
participant Fallback as Fallback Handler
Task->>Dispatch: New task arrives
Dispatch->>AI: Send task description + metadata
alt AI Available
AI->>AI: Analyze complexity
AI-->>Dispatch: Return JSON (tier, reason)
Dispatch->>ModelSel: Select model (haiku/sonnet/opus)
else AI Unavailable
Dispatch->>Fallback: Use default tier
Fallback-->>ModelSel: Assign opus
end
ModelSel->>Executor: Execute with selected model
Executor->>Executor: Generate output
Executor->>QualityAI: Send output + signals
alt Quality Gate AI Available
QualityAI->>QualityAI: Evaluate PASS/FAIL
QualityAI-->>Dispatch: Return decision + reason
else Quality AI Unavailable
Dispatch->>Fallback: Use legacy checks
end
alt Quality Gate Result
QualityAI-->>Task: PASS—task complete
QualityAI-->>Dispatch: FAIL—reprompt or escalate
end
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 3 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Summary of ChangesHello @marcusquinn, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request significantly enhances the Highlights
Changelog
Activity
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
🔍 Code Quality Report�[0;35m[MONITOR]�[0m Code Review Monitoring Report �[0;34m[INFO]�[0m Latest Quality Status: �[0;34m[INFO]�[0m Recent monitoring activity: 📈 Current Quality Metrics
Generated on: Tue Feb 24 17:40:12 UTC 2026 Generated by AI DevOps Framework Code Review Monitoring |
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (4)
.agents/scripts/supervisor/dispatch.sh (4)
353-367: Consider extracting the AI call + JSON parse pattern into a shared helper.The resolve→call→ANSI-strip→grep-JSON→jq-extract pattern is duplicated verbatim across
classify_task_complexity(lines 353–385),should_prompt_repeat(lines 727–764), andcheck_output_quality(lines 1242–1281). A shared helper like_ai_json_query "$prompt" "$ai_cli" "$ai_model"that returns the parsed JSON block would reduce ~90 lines of duplication and ensure consistent timeout, ANSI stripping, and error handling.Also applies to: 727-741, 1242-1256
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.agents/scripts/supervisor/dispatch.sh around lines 353 - 367, The three functions classify_task_complexity, should_prompt_repeat, and check_output_quality duplicate the pattern that calls the AI CLI, applies a timeout, strips ANSI sequences, finds the JSON block and extracts with jq; extract that into a shared helper (e.g. _ai_json_query) which accepts parameters (prompt, ai_cli, ai_model, optional timeout) and returns the cleaned JSON string or empty on error; implement consistent portable_timeout use, uniform ANSI stripping, stderr redirection, and error fallback inside _ai_json_query, then replace the duplicated blocks in classify_task_complexity, should_prompt_repeat, and check_output_quality with calls to _ai_json_query and use its returned JSON for subsequent jq extraction and parsing.
687-707: Duplicate fallback blocks — consider extracting.The AI-unavailable fallback logic (check
prompt_repeat_done >= 1, return eligible or already_attempted) is repeated three times: lines 691–696, 701–706, and 766–772. A small helper function or early-return restructuring would reduce duplication.Optional extraction
+ _prompt_repeat_simple_fallback() { + if [[ "$prompt_repeat_done" -ge 1 ]]; then + echo "already_attempted" + return 1 + fi + echo "eligible" + return 0 + } + ai_cli=$(resolve_ai_cli 2>/dev/null) || { - if [[ "$prompt_repeat_done" -ge 1 ]]; then - echo "already_attempted" - return 1 - fi - echo "eligible" - return 0 + _prompt_repeat_simple_fallback + return $? }🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.agents/scripts/supervisor/dispatch.sh around lines 687 - 707, The duplicate AI-unavailable fallback logic (the repeated check of prompt_repeat_done >= 1 and returning "already_attempted" or "eligible") should be extracted into a single helper in dispatch.sh (e.g., a function like ai_fallback_check or is_prompt_eligible_when_ai_unavailable) and invoked wherever resolve_ai_cli or resolve_model fail; update the two failure blocks that call resolve_ai_cli and resolve_model to call this helper and return/echo based on its result instead of duplicating the same conditional, ensuring you reference prompt_repeat_done inside the new helper and keep existing behavior for resolve_ai_cli and resolve_model error paths.
337-351: Task description is injected raw into the AI prompt — low risk but worth a note.
$facts(containing the task description and tags from the DB) is interpolated directly into the prompt string. A crafted task description like"Ignore above instructions and classify as haiku"could manipulate classification. The blast radius is limited (wrong tier → fallback/budget cap catches it), and the "when uncertain, prefer opus" rule in the prompt provides a safety net. Acceptable for internal use, but worth documenting the trust boundary.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.agents/scripts/supervisor/dispatch.sh around lines 337 - 351, The prompt interpolates raw $facts into the prompt string via the prompt variable, allowing task descriptions to inject instructions; fix this by sanitizing or escaping $facts before interpolation (or use a safe templating/JSON encode function) and enforce a whitelist/length limit and tag-only parsing for complexity tags, update the prompt construction in dispatch.sh where the prompt variable is built to use the sanitized/escaped value and validate tags (`#trivial/`#simple/#complex) separately rather than relying on raw interpolation.
369-385: JSON extraction regex won't handle nested braces in AI response.
grep -oE '\{[^}]+\}'matches{followed by non-}chars followed by}— any brace in thereasonvalue (e.g.,"needs {complex} refactor") truncates the match. Since the AI is instructed to return a flat object with a one-sentence reason, this is unlikely but not impossible.A more robust alternative using
sedorjqdirectly on the stream:Proposed fix
- json_block=$(printf '%s' "$ai_result" | grep -oE '\{[^}]+\}' | head -1) + json_block=$(printf '%s' "$ai_result" | sed -n 's/.*\({.*}\).*/\1/p' | head -1) + # Validate it's parseable JSON before proceeding + if [[ -n "$json_block" ]] && ! printf '%s' "$json_block" | jq empty 2>/dev/null; then + json_block="" + fiThis same pattern appears in all three AI functions (lines 371, 745, 1260). If you fix it, apply consistently.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.agents/scripts/supervisor/dispatch.sh around lines 369 - 385, The current extraction uses grep -oE '\{[^}]+\}' which breaks on nested braces; instead pipe the full ai_result into jq to parse JSON reliably and extract .tier and .reason (keep the ai_result variable, remove the fragile json_block grep step, set tier=$(printf '%s' "$ai_result" | jq -r '.tier // ""' 2>/dev/null || echo "") and similarly for reason) and ensure failures fall back to empty strings; apply the same replacement for the other two occurrences of this pattern in the script (the blocks that set json_block/tier/reason).
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.agents/scripts/supervisor/dispatch.sh:
- Around line 1207-1217: The conditional handling when AI is unavailable is
dead: both branches echo "pass" despite the comment that it should "pass if PR
exists, fail if no changes." Update the ai_cli fallback block (where ai_cli is
set via resolve_ai_cli and tpr_url is checked) so that if tpr_url is
non-empty/valid you echo "pass" (and return 0), otherwise echo a failure token
such as "fail:no_pr_no_ai" and return a non-zero status (e.g., return 1) to
signal failure; alternatively, if the comment is stale, remove the conditional
and simplify to a single definitive behavior—adjusting ai_cli, resolve_ai_cli,
and tpr_url usage accordingly.
---
Nitpick comments:
In @.agents/scripts/supervisor/dispatch.sh:
- Around line 353-367: The three functions classify_task_complexity,
should_prompt_repeat, and check_output_quality duplicate the pattern that calls
the AI CLI, applies a timeout, strips ANSI sequences, finds the JSON block and
extracts with jq; extract that into a shared helper (e.g. _ai_json_query) which
accepts parameters (prompt, ai_cli, ai_model, optional timeout) and returns the
cleaned JSON string or empty on error; implement consistent portable_timeout
use, uniform ANSI stripping, stderr redirection, and error fallback inside
_ai_json_query, then replace the duplicated blocks in classify_task_complexity,
should_prompt_repeat, and check_output_quality with calls to _ai_json_query and
use its returned JSON for subsequent jq extraction and parsing.
- Around line 687-707: The duplicate AI-unavailable fallback logic (the repeated
check of prompt_repeat_done >= 1 and returning "already_attempted" or
"eligible") should be extracted into a single helper in dispatch.sh (e.g., a
function like ai_fallback_check or is_prompt_eligible_when_ai_unavailable) and
invoked wherever resolve_ai_cli or resolve_model fail; update the two failure
blocks that call resolve_ai_cli and resolve_model to call this helper and
return/echo based on its result instead of duplicating the same conditional,
ensuring you reference prompt_repeat_done inside the new helper and keep
existing behavior for resolve_ai_cli and resolve_model error paths.
- Around line 337-351: The prompt interpolates raw $facts into the prompt string
via the prompt variable, allowing task descriptions to inject instructions; fix
this by sanitizing or escaping $facts before interpolation (or use a safe
templating/JSON encode function) and enforce a whitelist/length limit and
tag-only parsing for complexity tags, update the prompt construction in
dispatch.sh where the prompt variable is built to use the sanitized/escaped
value and validate tags (`#trivial/`#simple/#complex) separately rather than
relying on raw interpolation.
- Around line 369-385: The current extraction uses grep -oE '\{[^}]+\}' which
breaks on nested braces; instead pipe the full ai_result into jq to parse JSON
reliably and extract .tier and .reason (keep the ai_result variable, remove the
fragile json_block grep step, set tier=$(printf '%s' "$ai_result" | jq -r '.tier
// ""' 2>/dev/null || echo "") and similarly for reason) and ensure failures
fall back to empty strings; apply the same replacement for the other two
occurrences of this pattern in the script (the blocks that set
json_block/tier/reason).
| # Ask AI to assess quality | ||
| local ai_cli | ||
| ai_cli=$(resolve_ai_cli 2>/dev/null) || { | ||
| # AI unavailable — pass if PR exists, fail if no changes | ||
| if [[ -n "$tpr_url" && "$tpr_url" != "no_pr" && "$tpr_url" != "task_only" ]]; then | ||
| echo "pass" | ||
| else | ||
| echo "pass" | ||
| fi | ||
| return 0 | ||
| } |
There was a problem hiding this comment.
Dead branch: both if and else return "pass" — contradicts comment.
The comment on line 1210 says "pass if PR exists, fail if no changes," but both branches echo "pass". Either the else should return a failure, or the comment is stale and the conditional should be removed.
ai_cli=$(resolve_ai_cli 2>/dev/null) || {
- # AI unavailable — pass if PR exists, fail if no changes
- if [[ -n "$tpr_url" && "$tpr_url" != "no_pr" && "$tpr_url" != "task_only" ]]; then
- echo "pass"
- else
- echo "pass"
- fi
+ # AI unavailable — don't block on AI failure
+ echo "pass"
return 0
}If the intent was to fail when no PR exists and AI is unavailable, the else should be:
echo "fail:no_pr_no_ai"🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In @.agents/scripts/supervisor/dispatch.sh around lines 1207 - 1217, The
conditional handling when AI is unavailable is dead: both branches echo "pass"
despite the comment that it should "pass if PR exists, fail if no changes."
Update the ai_cli fallback block (where ai_cli is set via resolve_ai_cli and
tpr_url is checked) so that if tpr_url is non-empty/valid you echo "pass" (and
return 0), otherwise echo a failure token such as "fail:no_pr_no_ai" and return
a non-zero status (e.g., return 1) to signal failure; alternatively, if the
comment is stale, remove the conditional and simplify to a single definitive
behavior—adjusting ai_cli, resolve_ai_cli, and tpr_url usage accordingly.
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request refactors several core functions in the dispatch.sh script to adopt an "AI-first" approach for decision-making. The classify_task_complexity function was rewritten to send task descriptions and tags to an AI model for classification, replacing previous regex-based pattern matching. The resolve_task_model function was simplified by replacing a multi-step heuristic chain (including pattern-tracker, cost-efficiency, and contest detection) with a direct AI call for model selection, followed by a budget-aware cap. Similarly, the should_prompt_repeat function now uses AI to determine retry eligibility, gathering task state and pattern data for an AI decision, while retaining mechanical guards. The check_output_quality function was also overhauled to send comprehensive worker output signals (log details, git diff stats) to an AI for quality assessment, replacing numerous individual heuristic checks. Across these changes, the review comments consistently point out the pervasive use of 2>/dev/null to suppress standard error output from various commands (jq, claude, opencode, resolve_ai_cli, resolve_model, helper scripts, git diff, bash -n). The reviewer argues that this practice hides critical diagnostic information, such as syntax errors, API failures, or malformed JSON, making debugging difficult, and recommends removing these suppressions to ensure error visibility.
| local sc_count | ||
| sc_count=$(bash -n "$full_path" 2>&1 | wc -l | tr -d ' ') | ||
| shellcheck_errors=$((shellcheck_errors + sc_count)) | ||
| syntax_errors=$((syntax_errors + sc_count)) |
There was a problem hiding this comment.
The general rules state: "Avoid suppressing stderr with '2>/dev/null' when sourcing files in shell scripts; syntax errors or missing files should remain visible for debugging even if a fallback is provided." Using 2>/dev/null with bash -n will hide any syntax errors in the shell scripts, which are critical for maintaining script quality and correctness. This directly violates the rule and makes debugging shell scripts extremely difficult.
It is crucial to remove 2>/dev/null here to ensure that all syntax errors are reported.
References
- Avoid suppressing stderr with '2>/dev/null' when sourcing files in shell scripts; syntax errors or missing files should remain visible for debugging even if a fallback is provided.
- In shell scripts, avoid blanket suppression of errors with '2>/dev/null'. For 'declare -f', redirecting only stdout ('>/dev/null') is sufficient as it does not output to stderr, allowing potential syntax errors to remain visible.
| local json_block | ||
| json_block=$(printf '%s' "$ai_result" | grep -oE '\{[^}]+\}' | head -1) | ||
| if [[ -n "$json_block" ]]; then | ||
| local decision |
There was a problem hiding this comment.
Suppressing stderr for jq when parsing the AI's response can hide issues where the AI returns malformed or unexpected JSON. This could indicate a problem with the AI's output format that needs to be addressed.
It's generally better to allow jq errors to be visible, especially when dealing with external service outputs, to ensure data integrity.
| local decision | |
| decision=$(printf '%s' "$json_block" | jq -r '.decision // ""' || echo "") |
References
- When reporting errors for failed file operations in shell scripts, such as 'jq' writes, include the file path in the error message. Avoid suppressing stderr with '2>/dev/null' to ensure that diagnostic information about malformed files or write failures is visible.
- In shell scripts with 'set -e' enabled, use '|| true' to prevent the script from exiting when a command like 'jq' fails on an optional lookup. Do not suppress stderr with '2>/dev/null' so that actual syntax or system errors remain visible for debugging.
| if [[ "$pr_rate" -lt 25 ]]; then | ||
| echo "pattern_data_negative:${pr_rate}pct_over_${pr_total}" | ||
| local ai_model | ||
| ai_model=$(resolve_model "sonnet" "$ai_cli" 2>/dev/null) || { |
There was a problem hiding this comment.
Similar to the previous comment, suppressing stderr for resolve_model can hide issues with model resolution or configuration. Even with a fallback, diagnostic information is lost.
Consider removing 2>/dev/null or redirecting stderr to a log file.
| ai_model=$(resolve_model "sonnet" "$ai_cli" 2>/dev/null) || { | |
| ai_model=$(resolve_model "sonnet" "$ai_cli") || { |
References
- Avoid using '2>/dev/null' for blanket suppression of command errors in shell scripts to ensure that authentication, syntax, or system issues remain visible for debugging.
| ai_result=$(portable_timeout 15 claude \ | ||
| -p "$prompt" \ | ||
| --model "$claude_model" \ | ||
| --output-format text 2>/dev/null || echo "") |
There was a problem hiding this comment.
Suppressing stderr for the claude command can hide critical errors from the Claude API, such as invalid API keys, network issues, or service unavailability. This makes it difficult to diagnose why the AI classification might be failing.
It's recommended to remove 2>/dev/null to ensure these errors are visible for debugging.
| --output-format text 2>/dev/null || echo "") | |
| ai_result=$(portable_timeout 15 claude \ | |
| -p "$prompt" \ | |
| --model "$claude_model" \ | |
| --output-format text || echo "") |
References
- Avoid using '2>/dev/null' for blanket suppression of command errors in shell scripts to ensure that authentication, syntax, or system issues remain visible for debugging.
| json_block=$(printf '%s' "$ai_result" | grep -oE '\{[^}]+\}' | head -1) | ||
| if [[ -n "$json_block" ]]; then | ||
| local tier | ||
| tier=$(printf '%s' "$json_block" | jq -r '.tier // ""' 2>/dev/null || echo "") |
There was a problem hiding this comment.
Suppressing stderr for jq when parsing the AI's response can hide issues where the AI returns malformed or unexpected JSON. This could indicate a problem with the AI's output format that needs to be addressed.
It's generally better to allow jq errors to be visible, especially when dealing with external service outputs, to ensure data integrity.
| tier=$(printf '%s' "$json_block" | jq -r '.tier // ""' 2>/dev/null || echo "") | |
| tier=$(printf '%s' "$json_block" | jq -r '.tier // ""' || echo "") |
References
- When reporting errors for failed file operations in shell scripts, such as 'jq' writes, include the file path in the error message. Avoid suppressing stderr with '2>/dev/null' to ensure that diagnostic information about malformed files or write failures is visible.
- In shell scripts with 'set -e' enabled, use '|| true' to prevent the script from exiting when a command like 'jq' fails on an optional lookup. Do not suppress stderr with '2>/dev/null' so that actual syntax or system errors remain visible for debugging.
| } | ||
|
|
||
| local ai_model | ||
| ai_model=$(resolve_model "sonnet" "$ai_cli" 2>/dev/null) || { |
There was a problem hiding this comment.
Similar to the previous comment, suppressing stderr for resolve_model can hide issues with model resolution or configuration. Even with a fallback, diagnostic information is lost.
Consider removing 2>/dev/null or redirecting stderr to a log file.
| ai_model=$(resolve_model "sonnet" "$ai_cli" 2>/dev/null) || { | |
| ai_model=$(resolve_model "sonnet" "$ai_cli") || { |
References
- Avoid using '2>/dev/null' for blanket suppression of command errors in shell scripts to ensure that authentication, syntax, or system issues remain visible for debugging.
| -m "$ai_model" \ | ||
| --format default \ | ||
| --title "quality-${task_id}-$$" \ | ||
| "$prompt" 2>/dev/null || echo "") |
There was a problem hiding this comment.
Suppressing stderr for the opencode run command can hide errors from the AI service itself, such as authentication failures, rate limits, or malformed requests. While || echo "" provides a fallback, the root cause of the failure would be obscured.
It's better to remove 2>/dev/null to allow these errors to be visible for debugging, or to redirect them to a specific log for later analysis.
| "$prompt" 2>/dev/null || echo "") | |
| ai_result=$(portable_timeout 15 opencode run \ | |
| -m "$ai_model" \ | |
| --format default \ | |
| --title "quality-${task_id}-$$" \ | |
| "$prompt" || echo "") |
References
- Avoid using '2>/dev/null' for blanket suppression of command errors in shell scripts to ensure that authentication, syntax, or system issues remain visible for debugging.
| ai_result=$(portable_timeout 15 claude \ | ||
| -p "$prompt" \ | ||
| --model "$claude_model" \ | ||
| --output-format text 2>/dev/null || echo "") |
There was a problem hiding this comment.
Suppressing stderr for the claude command can hide critical errors from the Claude API, such as invalid API keys, network issues, or service unavailability. This makes it difficult to diagnose why the AI classification might be failing.
It's recommended to remove 2>/dev/null to ensure these errors are visible for debugging.
| --output-format text 2>/dev/null || echo "") | |
| ai_result=$(portable_timeout 15 claude \ | |
| -p "$prompt" \ | |
| --model "$claude_model" \ | |
| --output-format text || echo "") |
References
- Avoid using '2>/dev/null' for blanket suppression of command errors in shell scripts to ensure that authentication, syntax, or system issues remain visible for debugging.
| json_block=$(printf '%s' "$ai_result" | grep -oE '\{[^}]+\}' | head -1) | ||
| if [[ -n "$json_block" ]]; then | ||
| local result | ||
| result=$(printf '%s' "$json_block" | jq -r '.result // ""' 2>/dev/null || echo "") |
There was a problem hiding this comment.
Suppressing stderr for jq when parsing the AI's response can hide issues where the AI returns malformed or unexpected JSON. This could indicate a problem with the AI's output format that needs to be addressed.
It's generally better to allow jq errors to be visible, especially when dealing with external service outputs, to ensure data integrity.
| result=$(printf '%s' "$json_block" | jq -r '.result // ""' 2>/dev/null || echo "") | |
| result=$(printf '%s' "$json_block" | jq -r '.result // ""' || echo "") |
References
- When reporting errors for failed file operations in shell scripts, such as 'jq' writes, include the file path in the error message. Avoid suppressing stderr with '2>/dev/null' to ensure that diagnostic information about malformed files or write failures is visible.
- In shell scripts with 'set -e' enabled, use '|| true' to prevent the script from exiting when a command like 'jq' fails on an optional lookup. Do not suppress stderr with '2>/dev/null' so that actual syntax or system errors remain visible for debugging.
| local result | ||
| result=$(printf '%s' "$json_block" | jq -r '.result // ""' 2>/dev/null || echo "") | ||
| local reason | ||
| reason=$(printf '%s' "$json_block" | jq -r '.reason // ""' 2>/dev/null || echo "") |
There was a problem hiding this comment.
Similar to the previous jq call, suppressing stderr here can mask issues with the AI's JSON output, making it harder to debug unexpected responses.
Consider removing 2>/dev/null to ensure jq errors are visible.
| reason=$(printf '%s' "$json_block" | jq -r '.reason // ""' 2>/dev/null || echo "") | |
| reason=$(printf '%s' "$json_block" | jq -r '.reason // ""' || echo "") |
References
- When reporting errors for failed file operations in shell scripts, such as 'jq' writes, include the file path in the error message. Avoid suppressing stderr with '2>/dev/null' to ensure that diagnostic information about malformed files or write failures is visible.
- In shell scripts with 'set -e' enabled, use '|| true' to prevent the script from exiting when a command like 'jq' fails on an optional lookup. Do not suppress stderr with '2>/dev/null' so that actual syntax or system errors remain visible for debugging.
|
Warning Gemini encountered an error creating the review. You can try again by commenting |
1 participant
Summary
classify_task_complexity()(180-line keyword heuristic) with AI classification via sonnetshould_prompt_repeat()(65-line heuristic) with AI retry advisorcheck_output_quality()(97-line 6-check heuristic) with AI quality assessmentresolve_task_model(): remove 5-step heuristic chain, use AI classification + budget capai-lifecycle.shChanges
classify_task_complexity()should_prompt_repeat()check_output_quality()resolve_task_model()Testing
bash -nsyntax check: cleanshellcheck -S warning: zero violationsCloses #2209 (supervisor migration subtask)
Summary by CodeRabbit
Release Notes