Annotated Review

Annotated Review

Windows Runtime Commands

On native Windows, use the managed Windows runner and the native runtime command target. Set $runtime to the installed runtime root. Multi-agent installs usually use %LOCALAPPDATA%\ai-agents-skills\runtime. Then run:

$runtime = if ($env:AAS_RUNTIME_ROOT) { $env:AAS_RUNTIME_ROOT } else { "$env:LOCALAPPDATA\ai-agents-skills\runtime" }
& "$runtime\run_skill.ps1" "skills/annotated-review/review.py" <args>

POSIX examples below use run_skill.sh and .sh command targets; use the Windows command target above on native Windows.

This adapts the live OpenClaw paper-review workflow.

Trigger rule

Use this skill only when the request explicitly contains both ideas:

• "annotate" (or "annotation" / "annotated"), and
• "review"

Both ideas must be present in the user request. Review intent alone is not enough.

Examples that should trigger this skill:

• annotate and review this paper
• give me an annotated review
• annotate this paper, then review it
• annotate this paper and add the review to Zotero

Examples that should not trigger this skill by themselves:

• review this paper
• critique this paper
• hard review
• find issues in this paper
• multi-agent review
• review and add to Zotero

Routing rule for review-only requests:

• if the user asks only for a review, use the normal single-agent review flow via paper-review
• if the user asks for a multi-agent review, use agent-group-discuss
• do not auto-route review-only requests to annotated-review

Strict Zotero rule

Zotero is off by default for reviews.

• "Review this paper" -> do not use this skill automatically
• "Review and add to Zotero" -> still do not use this skill unless annotation is also requested

Do not touch Zotero unless the user explicitly asks.

Document lookup order for review tasks

If a review task requires locating the paper or book itself and the user did not already provide the source file/path, use this lookup order:

1. check the Zotero library
2. if not found there, check the Calibre library
3. only if neither library has the document, look online

For review tasks, do not go to online retrieval before checking both local libraries.

Base path

• $AAS_RUNTIME_WORKSPACE/skills/annotated-review/

Use the managed runtime runner rather than invoking run_review.sh directly. The runner sets the same workspace environment Claude uses.

Shared runner:

• bash "${AAS_RUNTIME_ROOT:-$HOME/.local/share/ai-agents-skills/runtime}/run_skill.sh"

Workflow imported from the bot

1. Review / annotate the paper.
2. Run an independent verification pass.
3. Run a trust-verification / citation-check pass.
4. Execute the review script against the paper or source tree.
5. If explicitly requested, store the resulting note in Zotero.

Codex adaptation

• Use spawn_agent for the independent verifier and trust-verifier when the task is large enough to benefit.
• Keep the main synthesis local.
• Use functions.exec_command for the live review scripts.
• If the paper/book is not already attached or given by path, route document lookup as: zotero -> calibre -> online fallback.
• Do not use this skill for review-only requests; reserve it for requests that explicitly mention both annotate/annotation and review.
• Before producing review prose or a stored note, load writing-style-settings.md and record the active style profile. For mathematical, TCS, graph-theoretic, Lean, or LaTeX manuscripts, also load math-manuscript-style.md.

Execution patterns

bash "${AAS_RUNTIME_ROOT:-$HOME/.local/share/ai-agents-skills/runtime}/run_skill.sh" skills/annotated-review/run_review.sh --precompile --source <path>

bash "${AAS_RUNTIME_ROOT:-$HOME/.local/share/ai-agents-skills/runtime}/run_skill.sh" skills/annotated-review/run_review.sh --review-file /tmp/review.json --pdf <file>

bash "${AAS_RUNTIME_ROOT:-$HOME/.local/share/ai-agents-skills/runtime}/run_skill.sh" skills/annotated-review/run_review.sh --review-file /tmp/review.json --source <dir> --zotero-key <key>

Output rule

Companion review artifacts are still useful even if LaTeX compilation fails. Report the best available artifact and any compile error explicitly. Final review or annotation artifacts should include style_profile_ref, policy_hash, active_overlays, active_requirement_ids, and style_applied; a bare style_applied: true value is not enough unless it is backed by the workflow's record of the loaded policy and selected requirements.

Recommended templates

When this skill is involved, consider these workflow templates (install via the workflow-templates artifact profile, or --with-deps to pull backing skills):

• cross-agent-adversarial-review -- Producer-never-confirmer adversarial review of a paper, proof, or code artifact across agent families with a fresh-agent confirmation gate.