Skip to content

docs: add Agent Profiles RFC#18

Draft
vincentkoc wants to merge 16 commits into
mainfrom
rfc/model-harness-profiles
Draft

docs: add Agent Profiles RFC#18
vincentkoc wants to merge 16 commits into
mainfrom
rfc/model-harness-profiles

Conversation

@vincentkoc

@vincentkoc vincentkoc commented Jun 17, 2026 •
edited by osolmaz
Loading

Copy link
Copy Markdown
Member

Status: In progress. A manual pass over the full RFC is still pending after the Schemator-guided naming update and Agent Profiles rename.

Summary

  • Proposes RFC 0009, Agent Profiles.
  • Replaces the binary local-model Lean direction with a versioned, deterministic profile registry.
  • Structures profiles with spec.common for harness-agnostic behavior and domain-owned sections such as spec.openclaw.ai for harness-specific behavior.
  • Separates model identity, Agent Profiles, provider drivers, and future managed-local Serving Presets.
  • Defines phase-one migration for Lean, GPT-5 response-style settings, and Claude/Opus thinking defaults, plus hosted-provider KV/cache non-regression boundaries.
  • Defines phase two as the ClawHub-backed Agent Profile distribution and discovery layer.

Scope

This is an RFC only. It does not change OpenClaw runtime behavior or add a configuration surface in this PR.

Verification

  • git diff --check
  • Targeted terminology search for stale pre-rename/model profile/profile/recipe naming
  • Manual review of RFC metadata, schema shape, phase boundaries, and PR-facing summary

RFC Lifecycle

  • Status: draft
  • Implementation issue: not created
  • Maintainer discussion thread: pending

@clawsweeper

clawsweeper Bot commented Jun 19, 2026 •
edited
Loading

Copy link
Copy Markdown

Codex review: found issues before merge. Reviewed July 1, 2026, 4:09 AM ET / 08:09 UTC.

Summary
The branch adds one draft Markdown RFC proposing Agent Profiles, including profile schema, selection, migration, Serving Presets, verification, and future ClawHub distribution.

Reproducibility: not applicable. this is an RFC/documentation PR, not a runtime bug report. The high-confidence checks are the PR diff, README lifecycle rules, current RFC inventory, and potential merge tree.

Review metrics: 3 noteworthy metrics.

  • RFC diff: 1 file added, 1046 additions, 0 deletions. This is a large product-design document, so manual RFC review is the main confidence gate.
  • Merge-tree RFC ids: 2 rfcs/0009-* files. The potential merge result would publish duplicate RFC ids unless this draft is renumbered.
  • Lifecycle state: draft status, blank issue, discussion pending. The repository lifecycle blocks merging draft RFCs before discussion and acceptance.

Merge readiness
Overall: 🦐 gold shrimp
Proof: 🌊 off-meta tidepool
Patch quality: 🦐 gold shrimp
Result: ready for maintainer review.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

  • Renumber the RFC to a free id and update self-references.
  • [P2] Add or link the maintainer-discussion thread before moving out of draft.
  • Update status and implementation issue only after maintainer acceptance.

Risk before merge

  • [P1] Merging as-is would publish two RFC documents numbered 0009, making references ambiguous.
  • [P1] The RFC is still marked draft with a blank implementation issue, and the PR body says the maintainer discussion thread is pending.
  • [P1] The RFC proposes new profile selection, configuration, migration, Serving Preset, and ClawHub distribution direction that needs maintainer product acceptance before publication.

Maintainer options:

  1. Finish RFC Lifecycle Before Merge (recommended)
    Renumber the RFC, link the maintainer discussion, and merge only after maintainer acceptance and implementation tracking are reflected in metadata.
  2. Keep Drafting In This PR
    Leave the PR open as a draft while naming, config scope, and ClawHub phase boundaries get the pending manual pass.
  3. Pause If Direction Is Unsponsored
    If maintainers do not want to sponsor Agent Profiles now, pause or close the draft rather than publishing unresolved RFC text.

Next step before merge

  • [P2] This member-authored draft RFC needs author and maintainer lifecycle/product action rather than an automated repair.

Security
Cleared: The diff adds one Markdown RFC and does not change executable code, workflows, dependencies, package metadata, secrets handling, or release scripts.

Review findings

  • [P2] Renumber this RFC before merge — rfcs/0009-agent-profiles.md:1-2
  • [P2] Keep the draft RFC out of main until lifecycle is complete — rfcs/0009-agent-profiles.md:8-10
Review details

Best possible solution:

Keep this as the active draft, assign a free RFC id, add or link the maintainer discussion, coordinate the ClawHub phase with #19, and update lifecycle metadata only after acceptance.

Do we have a high-confidence way to reproduce the issue?

Not applicable: this is an RFC/documentation PR, not a runtime bug report. The high-confidence checks are the PR diff, README lifecycle rules, current RFC inventory, and potential merge tree.

Is this the best way to solve the issue?

No, not as a merge candidate yet: the RFC is a reasonable vehicle for the product decision, but it needs renumbering, discussion linkage, lifecycle metadata, and maintainer acceptance before merge.

Full review comments:

  • [P2] Renumber this RFC before merge — rfcs/0009-agent-profiles.md:1-2
    The potential merge tree contains both current main's rfcs/0009-hosted-feeds-for-plugins-and-skills.md and this added rfcs/0009-agent-profiles.md, so landing would leave duplicate RFC ids. Pick a free id and update the filename plus self-references before merge.
    Confidence: 0.97
  • [P2] Keep the draft RFC out of main until lifecycle is complete — rfcs/0009-agent-profiles.md:8-10
    The README lifecycle says draft RFCs should not merge and each new RFC needs a maintainer-discussion thread. This file still has status: draft with a blank issue, and the PR body says the discussion thread is pending.
    Confidence: 0.94

Overall correctness: patch is incorrect
Overall confidence: 0.93

AGENTS.md: not found in the target repository.

Codex review notes: model internal, reasoning high; reviewed against a8e0ba2701d6.

Label changes

Label justifications:

  • P3: This is low-risk RFC documentation work with no immediate runtime behavior change.
  • merge-risk: 🚨 other: Merging as-is would publish duplicate RFC numbering and unresolved draft direction, which normal CI does not settle.
  • rating: 🦐 gold shrimp: Overall readiness is 🦐 gold shrimp; proof is 🌊 off-meta tidepool and patch quality is 🦐 gold shrimp.
  • status: ⏳ waiting on author: ClawSweeper has contributor-facing work open and is waiting for author action. Not applicable: This member-authored PR changes only RFC documentation, so external real behavior proof is not required.
Evidence reviewed

What I checked:

  • Target AGENTS.md check: No AGENTS.md exists under the target openclaw/rfcs git root, so no target-specific AGENTS policy applied. (a8e0ba2701d6)
  • Current main does not already contain Agent Profiles: A current-main search found no Agent Profiles, agentProfile, localModelLean, or profile-registry RFC text, so the central RFC is not already implemented in this repository. (a8e0ba2701d6)
  • Added RFC lifecycle state: The added RFC frontmatter is still draft with a blank implementation issue. (rfcs/0009-agent-profiles.md:8, 02df4b53ab14)
  • RFC lifecycle rule: The README says new RFCs should not merge while draft, need a maintainer-discussion thread, and should merge only after acceptance with an implementation issue. (README.md:80, a8e0ba2701d6)
  • Duplicate RFC id in merge tree: GitHub's potential merge tree contains both rfcs/0009-agent-profiles.md and rfcs/0009-hosted-feeds-for-plugins-and-skills.md. (10497e15ffb1)
  • Current main RFC 0009: Current main already has rfcs/0009-hosted-feeds-for-plugins-and-skills.md from the merged hosted-feeds RFC. (rfcs/0009-hosted-feeds-for-plugins-and-skills.md:1, a8e0ba2701d6)

Likely related people:

  • kevinlin-openai: git blame shows this person authored the README lifecycle rules that block merging draft RFCs before discussion, acceptance, and implementation tracking. (role: RFC process author; confidence: high; commits: e366ea9825a4, bbb4058da234, f4fdf38f4717; files: README.md, rfcs/0000-template.md)
  • Dallin Romney: Recent history shows this person clarified RFC sidecar and layout guidance in the README and template. (role: recent RFC layout contributor; confidence: medium; commits: 3aa7d727383f; files: README.md, rfcs/0000-template.md)
  • giodl73-repo: The merged hosted-feeds RFC on current main occupies RFC id 0009 and overlaps this proposal's future ClawHub distribution phase. (role: adjacent RFC author; confidence: medium; commits: 89cf322325b0, a8844eb53a2b; files: rfcs/0009-hosted-feeds-for-plugins-and-skills.md)
  • osolmaz: PR commits and discussion show this person drove the Agent Profiles rename/schema pass and noted that manual product-language review remains pending. (role: active RFC co-author and reviewer; confidence: high; commits: 0a41e99fb644, 8b4d35090cf1, 02df4b53ab14; files: rfcs/0009-agent-profiles.md)
What the crustacean ranks mean
  • 🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
  • 🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
  • 🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
  • 🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
  • 🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
  • 🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
  • 🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works
  • ClawSweeper keeps one durable marker-backed review comment per issue or PR.
  • Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
  • A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
  • PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
  • Maintainers can also comment @clawsweeper review to request a fresh review only.
  • Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
  • Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
  • Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

@osolmaz

osolmaz commented Jun 19, 2026 •
edited
Loading

Copy link
Copy Markdown
Member

Applied a Schemator-guided naming simplification pass to the model profile RFC. This kept the corrected stable vocabulary like extends, settings, toolExposure, contextPosture, and profile, while tightening names such as promptRecipe -> promptPreset, reasoningDefault -> reasoningMode, and the GPT setting shape around responseStyle.

Next I will do a manual pass over the full document to catch product-language issues, naming consistency, and any places where the automated schema review should not drive the final RFC wording.

@clawsweeper clawsweeper Bot added rating: 🐚 platinum hermit Good normal PR readiness with ordinary maintainer review expected. status: 👀 ready for maintainer look ClawSweeper has no concrete contributor-facing blocker left for this PR. rating: 🦐 gold shrimp Decent PR readiness signal, but merge confidence is limited. status: ⏳ waiting on author ClawSweeper has contributor-facing work open and is waiting for author action. P3 Low-risk cleanup, docs, polish, ergonomics, or speculative feature. merge-risk: 🚨 other 🚨 Merging this PR has meaningful risk outside the owned taxonomy. and removed rating: 🐚 platinum hermit Good normal PR readiness with ordinary maintainer review expected. status: 👀 ready for maintainer look ClawSweeper has no concrete contributor-facing blocker left for this PR. labels Jun 19, 2026
This was referenced Jun 20, 2026
Draft
Open
@osolmaz osolmaz changed the title RFC: Model Harness Profiles docs: add Agent Cards RFC Jun 22, 2026
@clawsweeper clawsweeper Bot added status: 👀 ready for maintainer look ClawSweeper has no concrete contributor-facing blocker left for this PR. and removed status: ⏳ waiting on author ClawSweeper has contributor-facing work open and is waiting for author action. labels Jun 22, 2026
@clawsweeper clawsweeper Bot removed rating: 🦐 gold shrimp Decent PR readiness signal, but merge confidence is limited. status: 👀 ready for maintainer look ClawSweeper has no concrete contributor-facing blocker left for this PR. labels Jun 22, 2026
@clawsweeper clawsweeper Bot mentioned this pull request Jun 22, 2026
Merged
@clawsweeper clawsweeper Bot added rating: 🦐 gold shrimp Decent PR readiness signal, but merge confidence is limited. status: 👀 ready for maintainer look ClawSweeper has no concrete contributor-facing blocker left for this PR. labels Jun 23, 2026
@osolmaz osolmaz changed the title docs: add Agent Cards RFC docs: add Agent Profiles RFC Jun 24, 2026
@clawsweeper clawsweeper Bot added status: ⏳ waiting on author ClawSweeper has contributor-facing work open and is waiting for author action. and removed status: 👀 ready for maintainer look ClawSweeper has no concrete contributor-facing blocker left for this PR. labels Jun 24, 2026
@clawsweeper

clawsweeper Bot commented Jul 22, 2026

Copy link
Copy Markdown

ClawSweeper status: review started.

I am starting a fresh review of this pull request: docs: add Agent Profiles RFC This is item 1/1 in the current shard. Shard 0/1.

This placeholder means the worker is alive and reading the current context. I will edit this same comment with the actual review when the claws are done clicking.

Crustacean status: shell secured, claws on keyboard, evidence pebbles being sorted.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

merge-risk: 🚨 other 🚨 Merging this PR has meaningful risk outside the owned taxonomy. P3 Low-risk cleanup, docs, polish, ergonomics, or speculative feature. rating: 🦐 gold shrimp Decent PR readiness signal, but merge confidence is limited. status: ⏳ waiting on author ClawSweeper has contributor-facing work open and is waiting for author action.

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@vincentkoc @osolmaz