Changelog

0.4.2

Remove opt-in --judge / network LLM second pass from the public product (offline-only surface).

Hermes Agent adapter (first-class): skill + pre_tool_call plugin + shell hook + AGENTS.md.
Installer resolves active Hermes profile via hermes config path (not only ~/.hermes).
Hermes skill frontmatter tags for Skills Hub / /promptguard discovery.
Docs: Hermes elevated alongside Claude/Codex/OpenClaw as a primary agent surface.
Installer: ./install-agent-adapters.sh hermes (also included in all).

OpenClaw plugin: PROMPTGUARD_PROFILE (default coding-agent) and PROMPTGUARD_FAIL_ON (default high); pack-aware rule filter.
Skill pack metadata under skills/promptguard/references/packs/.
Claude hook script uses package auditor with coding-agent profile.
promptguard tui PATH minimal interactive review (non-TTY table fallback).

Baseline regression: --baseline PATH and --fail-on-new on audit/audit-repo.
Match modes: word_any / word_also_any / word_missing_any and regex_* rule fields.
promptguard export-promptfoo converts eval JSONL to a promptfoo YAML skeleton.
Classic substring rule fields remain backward compatible.

Profiles: general, coding-agent, system, security (--profile).
Severity exit gate: --fail-on (legacy any-finding when omitted).
Rule packs under promptguard/packs/ plus --rules PATH.
Multi-surface extractors: SKILL/AGENTS/CLAUDE.md, JSON/YAML prompt keys, fenced blocks.
promptguard audit-repo with include/exclude globs.
SARIF 2.1.0 report format (--format sarif).
Grounded fix drafts (preserve original wording + contract checklist).
Accept-risk trail: --accept-risk ID:reason, --apply-accepted.
Security pack rules PG016–PG018 + eval/security_cases.jsonl.
Skill scripts delegate to package CLI for flag parity.