You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Auto-generated by compile-agentic-workflows.yml — these are what actually execute
Total
50
Verify with `ls .github/workflows/
Only the compiled .lock.yml files run. The .md sources are the source of truth and are reviewed in PRs; lock files are regenerated via the gh aw compile CLI.
Refresh every tracked data/cia/** + cia-data/** CSV from upstream Hack23/cia; refresh production-stats.json; inject counts into 14× index*.html; open a single PR on changes
🤖 Agentic News Generation (14 × 2 = 28 files)
Each agentic workflow is a pair: an authored .md source + a compiled .lock.yml. The source imports bounded-context prompt modules from .github/prompts/. The compiled lock file is hardened (SHA-pinned actions, egress firewall, five-layer safe-outputs) and is what GitHub Actions actually runs.
Translate analysis/daily/**/executive-brief.md into executive-brief_<lang>.md for the 13 non-English target languages (default batch: 3 sources × 13 languages = 39 files/run)
Every news workflow imports the bounded-context prompt modules in this exact order (full contract in .github/prompts/README.md):
Every news workflow declares the same tool & runtime surface for parity, resilience, and full gh-aw v0.74.3 capability coverage:
Field
Value
Purpose
runtimes.node.version
"26"
Pinned Node 26 for IMF CLI + render scripts
engine.id / engine.model
copilot / claude-sonnet-4.6
Faster Sonnet model (adopted in the v0.71.3 refactor for throughput within the 60-min budget; carried forward to v0.74.3 — GPT-5.4 / GPT-5.4-mini eligibility now resolved upstream per v0.74.3 release notes #31695 + #32197 but not yet A/B-tested on this repo)
engine.mcp.session-timeout
1hremoved
Was added in gh-aw v0.71.3 and rejected by MCP Gateway v0.3.1. v0.74.3 lock files now ship MCP Gateway v0.3.9; re-acceptance has not yet been validated on this repo, so the field stays removed until verified.
tools.github.toolsets
[all]
Full GitHub MCP surface (issues, PRs, repos, code-search, actions, releases, discussions, …); see github-tools.md
Full local tool surface; web-fetch reaches non-MCP public sources (statskontoret.se, riksdagsmonitor.com) through the AWF firewall
tools.cache-memory
keyed by news-${workflow}-${article_date}; best-effort cache persistence aligned with a 14-day recovery window
Resilience knob — analysis artifacts persisted at /tmp/gh-aw/cache-memory/; may be restored on the next run if the previous PR failed and the cache entry is still available (see 07-commit-and-pr.md §Cache-memory recovery)
tools.playwright
enabled in news-evening-analysis + news-realtime-monitor only
Live HTML validation for tier-C aggregation runs
features.mcp-gateway
true
Routes all MCP traffic through the gh-aw mcp-gateway (single audit point)
timeout-minutes
60
Job ceiling measured from job start; agent phases target completion by agent minute 40, PR by agent minute 42 (hard 45) to reserve setup/safe-output headroom
Ecosystem identifiers preferred per upstream network.md. The broad containers ecosystem (which would also permit ghcr.io, quay.io, gcr.io, mcr.microsoft.com, pkgs.k8s.io, …) is deliberately omitted to keep least-privilege egress; only the minimal Docker Hub hosts actually required to resolve node:26-alpine for the SCB and World Bank MCP servers are enumerated. Any future switch to ghcr.io, quay.io, or other registries must add the specific hosts and be reviewed against the egress allowlist policy before merge.
Least-privilege agent token; write capabilities live exclusively in the safe-outputs runner job
v0.74.3 capabilities — adoption status
Tracking grid for the features called out in the gh-aw v0.74.3 release notes. All 14 news-*.lock.yml files compile against v0.74.3 today; the rows below classify each new capability as adopted, planned (follow-up PR), or non-applicable.
v0.74.3 feature
Status
Notes
Glob patterns in add-labels safe outputs (#32022)
Planned
Replace hard-coded labels: [agentic-news, analysis-data] with agentic-* glob to allow agent-driven self-labelling of agentic-tier-c, agentic-long-horizon, agentic-election-cycle.
Issue Fields in safe outputs (#30846)
Planned
Move severity / horizon-band / election-cycle metadata off labels into issue fields on uptime-monitor and PIR roll-forward issues.
Verified clean: gh aw fix (dry-run) reports "No fixes needed" against all 14 news workflows on v0.74.3 (validated 2026-05-15). gh aw fix --list-codemods shows the highest registered codemod targeting our migration range introduced in v0.68.4 — there are no v0.71→v0.74 codemods that apply.
Warn-mode threat guardrails for safe outputs (#32399)
Planned
Tighten safe-outputs.threat-detection.continue-on-error: true → warn-mode with structured warning ingested into analysis-quality.md.
Agent compatibility matrix and validation (#32396)