Conversation
… safe output instructions - Add explicit "NEVER search for safe output tools via bash" instruction (matching working workflows) - Add MANDATORY PR Creation section with clear do/don't list - Fix weekday languages to respect input parameter (default: en,sv) instead of hardcoding all 14 - Add time-check bash helper before data gathering phase - Add explicit noop fallback when all queries return empty - Add CRITICAL FINAL REMINDER section at bottom with all safe output scenarios - Simplify verbose cross-referencing examples and data freshness JS code - Improve error handling table with concrete safe output actions Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>
…r, fix time-check fallback Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>
Copilot
AI
changed the title
[WIP] [aw] Fix News Evening Analysis workflow failure
fix: improve news-evening-analysis workflow reliability with stronger safe output guardrails
Mar 14, 2026
github-actions
bot
added
documentation
Contributor
🔍 Lighthouse Performance Audit
📥 Download full Lighthouse report Budget Compliance: Performance budgets enforced via |
Copilot AI
added a commit
that referenced
this pull request
Mar 14, 2026
1. Rename module from "AI pipeline" to "heuristic-based analysis pipeline"
to match the actual deterministic implementation (no LLM integration)
2. Gate Passes 2-3 on iterations >= 2 with stub fallbacks for iterations=1
3. Fix class docstring to match actual conditional behavior
4. Fix safety comment in generators.ts to reference AIAnalysisPipeline
5. Tighten hasCoalitionStress() — remove generic "motion"/"opposition"
keywords; require stronger conflict markers (avslag/reject/amendment/etc.)
6. Fix TAKEAWAY_PROP: remove broken %ss placeholder in French/Spanish;
use full plural word forms in all 14 language templates
7. Fix TAKEAWAY_BET: replace %sv combined placeholder with %verb key;
add betVerbForm() for language-correct verb conjugation
8. Fix TAKEAWAY_MOT: replace %sv combined placeholder with %verb key;
add motVerbForm() for language-correct verb conjugation
9. Fix TAKEAWAY_EU: remove suffix placeholders, use full plural forms
10. Fix CI: restore cross-referencing examples, filter patterns, and date
calculations in news-evening-analysis.md (regression from PR #1179 merge)
Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>
This was referenced Mar 14, 2026
Draft
Copilot AI
added a commit
that referenced
this pull request
Mar 14, 2026
Review comments addressed: 1. hasDomain() now works with all 14 languages via DOMAIN_NAME_TO_KEY reverse mapping 2. buildStakeholderSwot() localizes SWOT subject with requested lang 3. cacheKey() includes fullContent and notis lengths for collision resistance 4. PolicyDomain.key uses canonical domain keys (e.g. 'healthcare') not localized names 5. Added non-EN PESTLE and SWOT localization test assertions 6. buildPestleAnalysis() uses 'en' for internal trigger checks 7. Removed 'riksdag' from state-agencies signals (too generic) 8. Updated selectRelevantStakeholders docstring to list all 3 always-included groups 9. generateExecutiveSummary() uses fullText ?? fullContent for key passage 10. Replaced DOMAIN_EN_NAMES with canonical DOMAIN_NAME_TO_KEY from policy-analysis.ts CI fix: Updated agentic-workflow-mcp-queries.test.ts to match current workflow content after PR #1179 merged (removed Example 1-3 assertions, updated date calculation and cross-referencing pattern checks). Co-authored-by: pethers <1726836+pethers@users.noreply.github.com>
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
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.
The News Evening Analysis agentic workflow has been failing consistently since March 12 (runs #23064228552, #23017036455) — the agent completes execution without ever calling a safe output tool, resulting in empty
OUTPUT_TYPESand automatic failure.Comparison with the reliably-working
news-committee-reportsworkflow revealed several missing guardrails in the prompt.Missing safe output tool search prohibition
The working workflows include an explicit instruction preventing agents from wasting time searching the filesystem for tools. Added:
Weekday mode hardcoded 14 languages instead of respecting input
The prompt instructed the agent to manually generate HTML articles for all 14 languages (
[en, sv, da, no, fi, de, fr, es, nl, ar, he, ja, ko, zh]), contradicting the description ("core languages EN, SV") and the default inputen,sv. Translations are handled by the separatenews-translatedispatch workflow. Now resolvesLANG_ARGfrom thelanguagesinput parameter.Other improvements
create_pull_requestvsnoopSTART_TIMEvalidationsafeoutputs___noopimmediately when all MCP queries return emptyOriginal prompt
📱 Kick off Copilot coding agent tasks wherever you are with GitHub Mobile, available on iOS and Android.