Skip to content

v1.2.0 — Scoped Agent & Session-Based Learning (Tier 3)

Choose a tag to compare

@Ar9av Ar9av released this 28 Apr 03:32
· 1041 commits to main since this release

What's new

Scoped Agent

On UserPromptSubmit, Warden calls the Anthropic API (Haiku) to synthesise a minimal, task-specific rule set from the user's goal — restricting tools, file paths, and network access to only what the task genuinely requires. Falls back to keyword-based static heuristics when no API key is present. Rules persist in .prismor-warden/scoped/ for the session.

warden scope list
warden scope show --session-id <id>
warden scope clear <id>

Session-Based Learning

Mines historical session data for recurring uncovered command patterns, tracks false positives, and detects evasion attempts (e.g. backtick vs $() substitution bypassing rules).

warden learn                  # full analysis + candidate proposals
warden learn --candidates     # list pending rules
warden learn --apply <id>     # promote to policy.yaml
warden learn --reject <id>
warden learn --json

Security & correctness fixes (from review)

  • Prompt-injection mitigation: LLM-synthesised allowed_tools clamped to the known-good available set — a crafted task prompt can no longer expand the scoped policy
  • Command injection fix in warden scope edit: os.system → subprocess.run to prevent $EDITOR exploitation
  • KeyError: 'id' crash in warden learn output when candidates are freshly mined (not yet persisted)
  • Corrected misleading display text in the scoped-rules box
  • Removed dead get_scoped_dir() from store.py (path mismatch with scoped_agent)

Full changelog

See CHANGELOG.md for the complete entry.