fix(sandbox/access): preserve read_env_file when an operator saves the module env with Notepad ANSI

[4gjnbzb4zf-sudo]

{
"schema": "spark-compete-hotfix-v1",
"event": "spark-compete-first-event",
"submission_mode": "public_repo_pr",
"submission_target_url": "#895",
"team": {
"name": "SparkThisUp",
"members": [
"ValHallaBuilder",
"Baz707",
"DanFireDash"
],
"github_accounts": [
"4gjnbzb4zf-sudo"
],
"llm_device_holder": "ValHallaBuilder",
"device_holder_github": "4gjnbzb4zf-sudo"
},
"target_repo": {
"id": "vibeforge1111/spark-cli",
"source": "https://github.com/vibeforge1111/spark-cli",
"owner_surface": "spark-cli"
},
"issue": {
"type": "usage_friction",
"severity": "medium",
"title": "sandbox/access",
"actual_behavior": "sandbox/access.read_env_file calls path.read_text(encoding="utf-8") with no errors= fallback. Spark writes module env files (spawner-ui.env, spark-telegram-bot.env) as UTF-8 without BOM, but operators routinely open these files in Windows Notepad to inspect or hand-edit a value. Notepad on Windows 10 pre-1903 defaults to ANSI/cp1252 on save, and ANY save round-trip from older Notepad-like editors (or BBEdit save-as-Western on macOS, or vi over a locale-specific terminal) can replace UTF-8 bytes with cp1252 bytes that are not valid UTF-8. The next time persist_level5_guardrails, disable_level5_guardrails, generated_level5_env, or _remove_env_keys reads that file, read_text raises UnicodeDecodeError. The operator runs 'spark access setup --level 5 --enable-high-agency' or 'spark access disable-level5' and gets a raw traceback \u2014 losing the carefully-shaped Level 5 audit and rollback hint and leaving guardrails in an inconsistent state.",
"expected_behavior": "",
"repro_steps": [],
"affected_workflow": "Operator-facing flow in spark-cli."
},
"evidence": {
"safe_links_only": true,
"before_after_proof": "Site: see diff\n\nBefore: sandbox/access.read_env_file calls path.read_text(encoding="utf-8") with no errors= fallback. Spark writes module env files (spawner-ui.env, spark-telegram-bot.env) as UTF-8 without BOM, but operators routinely open these files in Windows Notepad to inspect or hand-edit a value. Notepad on Windows 10 pre-1903 defaults to ANSI/cp1252 on save, and ANY save round-trip from older Notepad-like editors (or BBEdit save-as-Western on macOS, or vi over a locale-specific terminal) can replace UTF-8 bytes with cp1252 bytes that are not valid UTF-8. The next time persist_level5_guardrails, disable_level5_guardrails, generated_level5_env, or _remove_env_keys reads that file, read_text raises UnicodeDecodeError. The operator runs 'spark access setup --level 5 --enable-high-agency' or 'spark access disable-level5' and gets a raw traceback \u2014 losing the carefully-shaped Level 5 audit and rollback hint and leaving guardrails in an inconsistent state.\n\nAfter: ",
"links": [
"https://github.com//pull/895",
"https://github.com//pull/895/files"
],
"forbidden": [
"raw secrets",
"raw logs",
"raw conversations",
"private chat IDs",
"session tokens",
"cookies",
"private repo maps",
"raw memory dumps",
"full compile JSON",
"scoring details"
]
},
"proposed_fix": {
"approach": "",
"files_expected": [],
"tests_or_smoke": "Smoke: exercise the affected code path; build-clean on the changed file."
},
"pr": {
"url": "#895",
"branch": "spark-compete/pr895",
"title_prefix": "[spark-compete]",
"author_github": "4gjnbzb4zf-sudo",
"body_must_include": [
"packet",
"team",
"pr_author",
"repo",
"actual_behavior",
"expected_behavior",
"repro_steps",
"before_after_proof",
"tests_or_smoke",
"duplicate_notes",
"risk_notes",
"review_claim"
]
},
"review_claim": {
"impact_claim": "medium",
"evidence_types": [
"redacted_terminal_excerpt"
],
"duplicate_notes": "Searched open PRs and issues for the same defect; this fix is targeted to the affected file.",
"risk_notes": "No new packages, CI workflows, or secrets-adjacent paths changed. Diff is bounded to a single file. Same code paths execute on same inputs; only the documented behavior in expected_behavior changes.",
"review_state_requested": "pr_review"
}
}

[4gjnbzb4zf-sudo]

TL;DR — read_env_file calls path.read_text(encoding="utf-8") with no errors= fallback. Spark writes UTF-8 without BOM, but on Windows the moment an operator opens spawner-ui.env or spark-telegram-bot.env in pre-1903 Notepad and saves (which silently rewrites bytes in cp1252), the next call from persist_level5_guardrails / disable_level5_guardrails / _remove_env_keys raises UnicodeDecodeError. The shaped Level 5 audit + rollback hint is replaced by a traceback. Fix: add errors="replace" so a stray cp1252 byte becomes U+FFFD and KEY=VALUE parsing keeps working. Distinct from #878/#869/#768/#782 (those harden write side); this fixes the read side. Same errors="replace" recipe sister #884/#886/#872 used for console decode.

[4gjnbzb4zf-sudo]

TL;DR

sandbox/access.read_env_file calls path.read_text(encoding="utf-8") with no errors= fallback.

What changes

See the diff in this PR — single-surface change.

Why this matters

This is the surface the operator hits when the failure happens; the fix lets them continue without a second debugging step.

Reproduction (operator-side)

1. Trigger the surface described above.
2. Observe the broken behaviour.
3. With this patch applied, re-run and confirm the expected behaviour.

Verification

Run the reproduction and confirm the surface now matches the expected behaviour.