# NousResearch/hermes-agent · Hermes agent rule enforcement and pre-action hooks

## Open now

- 5 issues, 0 PRs open on GitHub.
- #26742: \[Feature\]: First-Class Claim Verification & Audit Mechanism for Hermes Agent
- #32682: Skill triggering is purely prompt-driven with no code-level enforcement
- #40662: \[Feature\] PreToolUse enforcement hook — LLM doesn't follow system-prompt/memory rules under recency bias during debugging
- #49764: \[Bug\]: Hermes Agent fails to follow rules in memory and skills
- #68126: Framework limitation: agent cannot internalize repeated corrections — needs forced pre-action guardrails, not just system prompt rules
- _Broad area: open items may be separate problems._

These issues and PRs concern the Hermes agent's unreliable adherence to system-prompt rules (especially under recency bias), and requests or attempts to add code-level enforcement mechanisms such as PreToolUse hooks and pre-response validation rather than relying solely on prompt instructions.

## TL;DR

- 11 member(s): 5 open issue(s), 6 closed issue(s), 0 open PR(s), 0 merged PR(s), 0 PR(s) closed without merging.
- No fix found: no merged or attempted fix PR is on record for this cluster.
- **Recommended:** if this matches your symptom, consider adding to the existing issue(s) below rather than opening a new one.

## Maintainer resolution verdict

- Internal-tracker verdict over this cluster's issue members: `unresolved`. This is distinct from the fix status above. Weigh it yourself.

## How this cluster was grouped

- Link evidence (`LLM_ONLY`): no structural link between members; the grouping rests entirely on an automated relevance judgement over their text.
- Read the members yourself before treating them as one problem.
- Pairwise link density `0.09`: the share of member pairs that carry a link, counting a structural link in full and a relevance-only link at 0.36. `1.00` means every pair is linked.

## Open related issues

- #26742: \[Feature\]: First-Class Claim Verification & Audit Mechanism for Hermes Agent
- #32682: Skill triggering is purely prompt-driven with no code-level enforcement
- #40662: \[Feature\] PreToolUse enforcement hook — LLM doesn't follow system-prompt/memory rules under recency bias during debugging
- #49764: \[Bug\]: Hermes Agent fails to follow rules in memory and skills
- #68126: Framework limitation: agent cannot internalize repeated corrections — needs forced pre-action guardrails, not just system prompt rules

## Closed / duplicate issues

- #9534: Feature Request: Support for Pre-Action / Interceptor Hooks in Agent Lifecycle
- #22956: Feature request: Pre-response validation hook for capability claims
- #47446: title: "Feature Request: Agent-level pre-response hook for enforcing meta-workflow checks (e.g. skill pre-loading)" labels: \["enhancement", "agent-core"\]
- #60429: \[Bug\]: hermes  agent keeps violating rules
- #63278: Hermes is forgetful?!
- #63770: \[Feature\] PreToolUse enforcement hook — prompt-level rules unreliable under recency bias

## Fix status

- No merged fix on record.
- Subsystem: `agent`