Repository navigation
🧭 Idea Promotion Queue #763
Description
Generated 2026-10-19 by idea-triage from 92 open Ideas and 29 open epics. Advisory only: nothing here is approved.
Legend: 🟢 Ripe = concrete, unblocked, uncovered, so promote now · 🟡 Soon = one named gap away · ⚪ Not yet = vague, covered by an epic, or blocked.
🟢 Ripe to promote
| Idea | Why now | Pre-drafted promotion note |
|---|---|---|
#2349 ETag-cached conditional reads in gh_api_read |
Measured problem (5,700–6,900 req/h against a 5,000 limit, #2197). Small change in one shared helper. No dependency on the identity migration. | Add an on-runner ETag cache to gh_api_read so polling reads send If-None-Match and 304s skip the primary rate limit. Deliver it with before/after request-rate telemetry. |
| #2025 Short-lived GitHub App installation tokens | Has the most human engagement of the new ideas (3 comments). Addresses the same quota and credential-blast-radius pain. Standard actions/create-github-app-token pattern. |
Replace GH_PAT_* cross-repo operations with per-run App installation tokens, one workflow at a time. Start with the canary ring and keep the PAT as a fallback until the stable ring is clean. Overlaps the "own identity" pillar of #2226, so scope it as a standalone step. |
| #2350 Per-PR attempt ledger with no-progress detection | Targets observed bot ping-pong and repeated refused passes. Pure, testable check. Complements the daily pr-runaway-detect.sh. |
Record each dev-lead pass (head SHA, intent, finding id, outcome) and stop and escalate once when a pass would repeat the same action with the same result. |
| #2352 Workflow lockfiles on the reusable-workflow side | Time-sensitive: public preview began 14 Oct. Scope is already bounded to reusable workflows, which leaves the channel-tag stubs of ADR-0002 alone. | Onboard .github-private's reusable workflows to GitHub workflow lockfiles. Lock third-party actions and their transitive dependencies, add a repo-jacking check in CI, and leave caller stubs out. |
🟡 Soon
| Idea | Missing before ripe |
|---|---|
| #2351 Intent-scoped write fences | No engagement yet. Needs a decision on how the allowed-path set is derived. Best sequenced after #2350 so both share the dev-lead pass record. |
| #2024 Independent disposition verification | Needs a per-role list of verifiable artifacts. Overlaps the run-contract work in #2226. |
| #1939 Merge queue readiness | Valuable, but it is blocked on the five prerequisites in #1871. Check for overlap with the approved #735. |
| #2098 Sandboxed agent Bash with egress allowlist | Needs a spike on sandbox support on hosted runners and the action tiers. |
| #2023 Deadlock root-cause attribution | No engagement. Needs a concrete list of gates and a sample of past stalls. |
| #2026 Agent tooling supply-chain verification | Overlaps #2352. Re-assess once lockfiles land. |
| #1842 Runtime workflow self-identity | Depends on the ADR-0007 ingress-collapse rollout (#2177) settling. |
⚪ Not yet / covered
- #2097 Model catalog sync: covered by epic Haiku 5.5 triage upgrade by hand, then fully automated model onboarding (detect → price → prove → shadow → promote → watch, Phases A–E) #2113 (Phase A, [Phase A] Daily model detection job that diffs provider model lists against the configured chains and opens one onboarding record per new model #2125; Phase E, [Phase E] Retirement watcher: start the successor's onboarding, with a deadline, when a configured model gets a retirement date #2130).
- #2226 Factory MVP: too broad and still tied to the proposed ADR-0013. Split it and decide first. Parts overlap 💡 Short-Lived GitHub App Installation Tokens for Cross-Repo Agent Operations #2025 and 💡 Independent Agent Disposition Verification to Catch False Success Claims #2024.
- #2242 Situation-room view: research verdict only, no deliverable.
- #2134, #1348, #2012: research or evaluation threads with no defined deliverable. Integrate PR Review into Dev Lead CLI #2012 could be scoped as a pilot.
- #1941, #564: complexity-aware model routing. Wait for the Haiku 5.5 triage upgrade by hand, then fully automated model onboarding (detect → price → prove → shadow → promote → watch, Phases A–E) #2113 rollout automation.
- #1940: the intro pricing window ends in December. Needs an eval-evidence check first.
- #792: stacked PRs are still in preview.
- #1943, #1942, #636, #794, #1841, #1840, #1839, #1838: no human engagement, so wait for signal. 💡 OWASP Agentic Top 10 Security Posture Scoring for the Agent Fleet #636 and 💡 OWASP Agentic Top 10 Posture Scoring: Operationalize the Published Framework #794 are duplicates of 💡 OWASP Agentic Top 10 Posture Scoring with Incident-Validated Threat Model #1942.
- Older un-engaged 💡 ideas from June to July (💡 Fleet-Wide Action Version Currency Probe with checkout v7 Backport Readiness #1155–💡 Canary Burn-Rate SLO with Post-Promotion Health Monitoring and Auto-Rollback #1158, 💡 Agent Session Telemetry Dashboard for Fleet-Wide Runtime Analytics #1040–💡 Eval-Score-Gated Canary Promotion for Quality-Aware Progressive Delivery #1042, 💡 Stale-Base Semantic Revert Detection Guard #790, 💡 Agent Pipeline Stall Detector & Automatic Recovery Escalation #791, 💡 OpenTelemetry GenAI Semantic Conventions for Agent Harness Observability #634, 💡 Structured Agent Decision Audit Trail with Provenance Export #637, 💡 GitHub Models API Multi-Model Chains for the Copilot Engine Tier #642, 💡 Anthropic Usage & Cost API Integration for Ground-Truth Budget Tracking #643, 💡 AWI Event-Context Input Sanitization for Agent Prompt Injection Defense #651, 💡 Effective Token (ET) Anomaly Alerting with Runaway Session Detection #565, 💡 Shadow Mode Dual-Run for Safe Agent Canary Validation #566, 💡 LSP-Grounded Finding Verification for PR Review False-Positive Reduction #594–💡 Token Cost Observatory LSP Metrics: Measuring the ROI of Semantic Code Intelligence #596, and similar): stale with no engagement. Consider closing or merging them.
- #973 and #958: TEST discussions, so close them.
✅ Already approved (not re-ranked)
#2095, #1893, #1711, #1400, #1360, #1043, #1039, #962, #933, #906, #877, #866, #793, #735, #666, #656, #653, #650, #641, #639, #607, #593, #578, #572, #563, #1118
To promote an idea, add the idea:approved label to its Discussion. That fires the BMAD Scrum Master initiative-planner, which plans the epic. This queue never approves anything.
Activity