Stage 3 executors and logs lacked structured access to ADO build vars — particularly the BUILD_TRIGGEREDBY_* set surfaced by the pipeline-completion trigger IR — so add_build_tag was reaching into std::env::var directly to fill the gap.

Summary

• ExecutionContext (src/safeoutputs/result.rs) — adds 14 Option<…> fields covering build identity (build_id: Option<u64>, build_number, build_reason, definition_name, source_branch, source_branch_name, source_version), ResourceTrigger upstream context (triggered_by_build_id, triggered_by_definition_name, triggered_by_build_number, triggered_by_project_id), and PR context (pull_request_id, pull_request_source_branch, pull_request_target_branch).
• from_env_lookup<F> constructor — Default::default() now delegates to a generic env lookup so tests drive population from a HashMap-backed closure rather than mutating process-global env (which is unsafe in Rust 2024 and the repo avoids it).
• add_build_tag.rs — replaces the inline std::env::var("BUILD_BUILDID") with ctx.build_id, eliminating the last executor that read build vars directly from env.
• log_execution_context (src/execute.rs) — emits build_id, build_reason, and triggered_by_definition_name so trigger source is visible in Stage 3 logs.
• Test ergonomics — converts existing ExecutionContext { ... } literals across the executor tests to ..Default::default() so future field additions don't ripple through unrelated test files.

Fields not yet read by production code carry #[allow(dead_code)]; they're populated for executor consumers added in follow-ups. No template change is required — ADO auto-exposes all BUILD_*/SYSTEM_* predefined variables to bash steps; only SYSTEM_ACCESSTOKEN (already wired via executor_ado_env) needs explicit mapping. CLI overrides and Stage 1 exposure are intentionally deferred per the plan.

// New testing pattern — no env mutation required
let ctx = ExecutionContext::from_env_lookup(env_from(&[
    ("BUILD_REASON", "ResourceTrigger"),
    ("BUILD_TRIGGEREDBY_DEFINITIONNAME", "Upstream Build"),
]));
assert_eq!(ctx.triggered_by_definition_name.as_deref(), Some("Upstream Build"));

Test plan

• cargo build, cargo clippy --all-targets --all-features — clean (no new warnings).
• cargo test — 1057 passing, including 8 new tests in safeoutputs::result covering field population, build_id numeric/non-numeric/absent parsing, and absent triggered_by_*/pull_request_* fields.