Issue Station
Sign In
Tools hereToolbox

hermes - ✅(Solved) Fix docs(curator): 'agent-created' definition is stale — docs claim foreground-created skills are curated, but PR #19621 restricted to background-review only [1 pull requests, 1 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
NousResearch/hermes-agent#29017•Fetched 2026-05-20 04:00:36
View on GitHub
Comments
0
Participants
1
Timeline
5
Reactions
0
Author
TwanislasTwanislas
Participants
TwanislasTwanislas
Timeline (top)
labeled ×3cross-referenced ×2

Error Message

Users see hermes curator status reporting zero agent-created skills despite having skills created by the agent in foreground sessions. The report shows Model: (not resolved) via (not resolved) with Duration: 0s, which looks like a configuration error but is actually expected behavior (LLM pass skipped because no candidates exist).

Root Cause

Users see hermes curator status reporting zero agent-created skills despite having skills created by the agent in foreground sessions. The report shows Model: (not resolved) via (not resolved) with Duration: 0s, which looks like a configuration error but is actually expected behavior (LLM pass skipped because no candidates exist).

Fix Action

Fixed

PR fix notes

PR #29018: docs(curator): align 'agent-created' definition with actual provenance semantics

Description (problem / solution / changelog)

Summary

The curator docs stated that any skill not bundled/hub-installed was "agent-created" — including foreground-created skills and hand-written ones. Since PR #19621, the curator requires an explicit created_by: "agent" marker in .usage.json, which only the background self-improvement review fork sets.

Closes #29017.

Changes

1. Rewrite "What agent-created means" section

Before (stale): claimed all non-bundled/non-hub skills were curated — including foreground skill_manage(create) calls and hand-written skills.

After: documents the actual 3-step eligibility check:

  1. Not bundled
  2. Not hub-installed
  3. created_by: "agent" or agent_created: true in .usage.json

Explains that only the background review fork sets this marker, and that foreground-created skills are intentionally excluded (user-directed by design). Warns that hand-written skills are NOT curated.

2. Add note on (not resolved) display in reports

When no agent-created candidates exist, the LLM pass is skipped and the report shows Model: (not resolved) via (not resolved) with Duration: 0s. This is expected behavior, not a config error. Added a :::note callout in the Per-run reports section.

3. Link to provenance internals

References tools/skill_provenance.py (write-origin ContextVar) and mark_agent_created() so users who want to understand the mechanism can trace the code.

Files changed

  • website/docs/user-guide/features/curator.md: +45 −21 lines

Verification

No code changes — docs only. The provenance semantics described match the current behavior in:

  • tools/skill_provenance.py — BACKGROUND_REVIEW ContextVar, is_background_review()
  • tools/skill_manager_tool.py — mark_agent_created() gate at line 782
  • tools/skill_usage.py — _is_curator_managed_record(), list_agent_created_skill_names()

Changed files

  • website/docs/user-guide/features/curator.md (modified, +45/-21)
RAW_BUFFERClick to expand / collapse

Problem

The Curator docs state:

"Everything else in ~/.hermes/skills/ is fair game for the curator. This includes: Skills the agent saved via skill_manage(action="create") during a conversation."

This is no longer true since PR #19621 (May 4, 2026), which restricted the mark_agent_created() call to the background self-improvement review fork only. The commit message is explicit:

"skills a user asks a foreground agent to write via skill_manage(create) now stay invisible to the curator. Only skills the background self-improvement review fork sediments through skill_manage get the created_by=agent marker."

The actual eligibility check now requires three conditions (not just two):

  1. Not in .bundled_manifest
  2. Not in .hub/lock.json
  3. Has "created_by": "agent" or "agent_created": true in .usage.json

Only the background review fork sets this marker — foreground skill_manage(action="create") calls do not.

Impact

Users see hermes curator status reporting zero agent-created skills despite having skills created by the agent in foreground sessions. The report shows Model: (not resolved) via (not resolved) with Duration: 0s, which looks like a configuration error but is actually expected behavior (LLM pass skipped because no candidates exist).

Proposed Fix

PR attached: update the "What agent-created means" section to accurately document the 3-step eligibility check, the foreground/background distinction, and the (not resolved) display in reports.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#autograd error#model save/load#optimization#mixed precision#configuration error

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING