feat(scenario): strict-toolban-fidelity-under-pressure — procedural compliance with harasser (BFCL V4 / τ²-bench / IFBench anchor)

feat(scenario): proc-only-build-no-combat — RBAC role separation under visible threat (BFCL V4 anchor)

feat(scenario): mcv-deploy-near-resource — econ-first siting near ore patch (SC2LE econ / Weber anchor)

feat(scenario): proc-no-attack-passive-only — ROE-style recon-only compliance (BFCL V4 / IFBench anchor)

feat(scenario): mcv-deploy-relocate-under-pressure — BCP relocation under attack (PlanBench replanning anchor)

feat(scenario): mcv-deploy-second-base — multi-site ramp-up via 2nd MCV (SC2LE 2-base macro anchor)

feat(scenario): proc-ordered-action-strict — strict ordered placement via then-composite (PlanBench / IFBench anchor)

feat(scenario): mid-economy-under-fire — live ore-patch defense vs raider bot (SC2LE / MineRL anchor)

feat(scenario): mcv-deploy-and-build — deploy MCV then build basic base (SC2LE BuildOrder anchor)

feat(scenarios): merge PR #8 from yiyu-tian — 4 new packs (rebased onto Wave 1+2)

feat(engine+bench): MCV deploy works for scenario-declared actors

feat(bench): then composite predicate happened-before (PlanBench replanning / PERT anchor)

feat(engine+bench): raider bot — worker-priority harasser (Rust ScriptedBehavior + bench registration)

feat(bench): forbidden_tools + tool_violations_gte — strict-toolban / procedural-compliance primitive (BFCL V4 / τ²-bench / IFBench anchor)

docs: Wave 1 engine-footgun aggregation in CLAUDE.md + regen scenarios.html

feat(scenario): perception-count-the-threat — exact-count state estimation (ERQA / POMDP anchor)

feat(scenario): mid-concede-vs-hold — abandon one base, save the other (PlanBench / CICERO retreat anchor)

feat(scenario): rebuild economy-harvest-investment — un-quarantine post-S0/S1 (PlanBench reinvestment anchor)

feat(scenario): navigation-confined-hard-only — confined egress under clock (ERQA / AI2-THOR anchor)

feat(scenario): harass-response-preserve — probe in your base, don't bleed (τ²-bench distractor anchor)

feat(scenario): defense-rush-survive — hold the line opening (MicroRTS rush-defense / SC2LE tempo anchor)

feat(scenario): strategy-trilemma — expand vs tech vs army (PlanBench / lmgame-Bench resource-allocation anchor)

feat(scenario): coordination-staggered-window — true parallel scheduling (Watch-And-Help / SMAC anchor)

feat(scenario): rebuild economy-harvest-timebox — un-quarantine post-S0/S1 (SC2LE CollectMinerals anchor)

feat(scenario): tempo-double-window — strike, lull, strike (TextStarCraft II / PlanBench tempo anchor)

schema: add meta.benchmark_anchor (list[str]); + SCENARIO_BACKLOG.md (200-pack catalog)

docs: CLAUDE.md + AGENTS.md — pointer for friend's Claude Code / Codex agents to use the no-cheat scenario methodology

docs: whole-suite no-cheat pass summary + engine footguns; regenerate catalog

strategy-gauntlet: no-cheat redesign — defended corridor + reachable timeout fail (recovered from stalled #18 agent)

rush-hour: no-cheat redesign — coverage+kill sweep beats auto-battle blob

strategy-twobody: no-cheat redesign — two enemy bases + region-scoped destroy predicate; simultaneous two-group control is now load-bearing

strict-production-bom: no-cheat redesign — reachable deadline, hard solvable, no stall-draws

perception-frontier-reading: no-cheat redesign — buildings_discovered_gte + reachable budgets so blind/stall LOSE, true frontier read WINS

building-and-planning: no-cheat redesign — engine-enforced tech chain + reachable timeout LOSS

perception-target-vs-fog: no-cheat redesign — reachable clock + fail, blind/stall LOSE, scout-then-commit WINS

reasoning-frontier-commit: no-cheat redesign — deadline now bites, lazy/brute/wrong-commit all LOSE

economy-force-buildup: no-cheat redesign — budget buys a 2nd powr + e3 force, spam loses

economy-investment: no-cheat redesign — tight indivisible budget makes wide/deep mutually exclusive

strict-sequence: no-cheat redesign — latched waypoint_sequence + reachable timeout LOSS

custom-map-no-enemy: no-cheat redesign — tight reachable deadline + real LOSS fail

#7 artofwar-sequenced-citadel: no-cheat redesign — ordered chain enforced

#6 artofwar-lure-the-tiger: adopt rigorously-verified no-cheat redesign

#6 artofwar-lure-the-tiger: no-cheat redesign — lure (timing) enforced

#5 artofwar-indirect-approach: no-cheat redesign — hazard-detour enforced

#4 artofwar-decoy-sacrifice: no-cheat redesign — decoy-sacrifice enforced

#4 artofwar-decoy-sacrifice: structural fixes (deadline, marker, off-map)

#3 adversarial-duel: realign inert deadlines + reachable timeout fail

fix-forward: repair merged PR#5/#6 (unblocked by own-unit actor_type fix)

Merge pull request #6 from yiyu-tian/yiyu/scenarios-fail-conditions-longhorizon

Merge pull request #5 from yiyu-tian/scenarios/tech-production-planning