fix(save_env_value): quote values with internal spaces to prevent shell-sourcing breakage (#66482)

[webtecnica]

Summary

One-line fix: extend the needs_quoting trigger in _quote_env_value() to also fire when any character in the value is whitespace (any(c.isspace() for c in value)).

Root cause

needs_quoting checked for #, quotes, and leading/trailing whitespace — but not internal whitespace. A path like /Users/paulo/Library/Application Support/hermes/keys/id_ed25519 was written unquoted, which breaks set -a; . .env (POSIX word-splitting).

Changes

• hermes_cli/config.py line 7781: added or any(c.isspace() for c in value) to the needs_quoting condition
• Escaping/wrapping is unchanged (existing replace('\\', ...) / f'...' handles the quoting correctly)

Closes #66482

[alt-glitch]

This was generated by AI during triage.

Duplicate of #66483, which already applies the same save_env_value whitespace-quoting repair with broader regression coverage.

[tonydwb]

Code Review Summary

Verdict: Comment

Critical

• None

Assessment

Adds a created_by != "agent" guard to the background skill curator. Prevents the autonomous LLM consolidation pass from archiving manually authored skills (e.g. installed via URL or direct SKILL.md authoring). Correctly skips skills without the created_by: "agent" marker.

Looks Good

• Clean guard using skill_usage._is_curator_managed_record
• Error message is clear and actionable
• No impact on agent-created skills

---

Reviewed by Hermes Agent

[teknium1]

Thanks for the focused serializer fix. The current predicate still omits internal whitespace (hermes_cli/config.py:7776-7781), while save_env_value writes that helper's result on both update and append paths (hermes_cli/config.py:7827-7841), so the premise holds on current main.

Problems

• This PR changes the serialization contract without adding regression coverage. The adjacent tests cover hash/quote/backslash cases (tests/hermes_cli/test_config.py:561-608) but not an internally spaced value.

Suggested changes

• Add a focused test that writes an internal-whitespace value, checks it is emitted as a double-quoted .env assignment, and confirms load_env() reads the original value back.

Automated hermes-sweeper review.

[teknium1]

Please add a regression test for an internal-space or tab value. Nearby coverage verifies hash/quote escaping, but not this new quoting trigger.

[OutThisLife]

Superseded by #67192 for #66482 (salvages @pnascimento75's #66483 with a round-trip regression test). Closing as redundant. Thanks @webtecnica.