agentsclimarketplace

Reconciliation Subagent Skill

Skill Preciso-GR/skills_for_preciso/Reconciliation-Subagent-Skill

Community library of agent extraction skills for Preciso GraphRAG — markdown SKILL.md files that teach coding agents (Claude Code, Codex, Copilot) to turn documents into graph-ready JSON.

Install
npx -y skills add Preciso-GR/skills_for_preciso --skill Reconciliation-Subagent-Skill

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use this skill when you are a reconciliation subagent asked to clean a single extraction JSON (not raw document text). Your job is to identify entity name variants, duplicate relationships, conflicts, or orphaned references and produce a patch file. Do NOT call any ingest tools. Do NOT rewrite the base extraction. Output must be a patch JSON file with merge/remove/flag directives only.

SKILL.md

2.7 KB, 559 tokens by cl100k_base, as published. Nobody here has run it

Reconciliation Subagent Skill

Agent Runtime Contract

Use this skill only after an extraction JSON already exists in extractions/.

Execution contract:

  • Read the extraction JSON, not the original raw document in to_be_extracted/.
  • Write only patch directives.
  • Do not replace the base extraction file.
  • Return control to the main agent so it can ingest the base extraction plus patch flow afterward.

Purpose

You are a reconciliation subagent. You do data cleanup on an existing extraction JSON. You are not extracting from the original document. Your output is a patch file that the main agent applies before ingestion.

Inputs You May Receive

  • Entities list only
  • Relationships list only
  • Entities + relationships together

You will NOT receive the full document text.

Output Patch Schema (Required)

Write a patch file with one or more of these keys:

{
  "merge_entities": [
    {
      "keep": "APPLE INC",
      "remove": ["APPLE", "APPLE INC."],
      "reason": "same company, variant names"
    }
  ],
  "remove_relationships": [
    {"src_id": "APPLE INC", "tgt_id": "NET_SALES_FY2024", "keyword": "REPORTED_METRIC"}
  ],
  "flag_conflicts": [
    {
      "type": "metric_conflict",
      "details": "Net sales reported as both +2% and -1% in FY2024",
      "entities": ["NET_SALES_FY2024"],
      "relationships": [
        {"src_id": "APPLE INC", "tgt_id": "NET_SALES_FY2024", "keyword": "REPORTED_METRIC"}
      ]
    }
  ],
  "broken_relationships": [
    {"src_id": "APPLE INC", "tgt_id": "MISSING_ENTITY", "keyword": "OWNS"}
  ],
  "suggested_fixes": [
    "Consider adding entity MISSING_ENTITY or removing the relationship."
  ]
}

Rules

  • You are cleaning data, not extracting from documents.
  • Be conservative. Only merge entities if you are certain they refer to the same real-world entity.
  • For temporal entities (FY2024, Q3 2024), NEVER merge even if similar. Different periods are different entities.
  • Use the exact entity_name values you see in the extraction.
  • Do not add new entities or relationships. Only rename, remove, or flag.
  • Write the patch file to the filename you were given and stop.
  • Do not call any ingest tool.

If You Are Unsure

If something is ambiguous, add it to flag_conflicts instead of merging.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.