The 5 LPA-review agents
/fundadmin:fund review-lpa reads the LPA and classifies the fund. Four agents (terms, risks, compliance, obligations) then run in parallel. The recommendations agent runs after all four return, because it builds on their findings. The five results are weighted into the LPA Safety Score, and the report is written once, complete.
The weights
The ring is the plugin's mark: one circle split in the five weights, the risk arc in teal.
- Terms & Provisions · 20%
fundadmin:fund-terms - Risk Assessment · 25%
fundadmin:fund-risks - Regulatory Compliance · 20%
fundadmin:fund-compliance - Obligations & Timeline · 15%
fundadmin:fund-obligations - Recommendations · 20%
fundadmin:fund-recommendations
| Agent | Dimension | Weight | What it does |
|---|---|---|---|
fundadmin:fund-terms | Terms & Provisions | 20% | Extract and categorize every LPA provision; its output feeds the other four agents |
fundadmin:fund-risks | Risk Assessment | 25% | Score each provision 1-10 for GP-favorability, LP exposure and deviation from market |
fundadmin:fund-compliance | Regulatory Compliance | 20% | Check an LPA against SEC, ERISA, FATCA/CRS, AML/KYC, AIFMD and SFDR rules |
fundadmin:fund-obligations | Obligations & Timeline | 15% | Map every obligation, deadline, notice and cure period into a timeline |
fundadmin:fund-recommendations | Recommendations | 20% | Turns the other four outputs into ranked negotiation points and language |
The score
| Score | Grade | Label |
|---|---|---|
| 90-100 | A+ | LP-Favorable |
| 80-89 | A | Balanced |
| 70-79 | B | Acceptable |
| 60-69 | C | Caution |
| 40-59 | D | GP-Heavy |
| 0-39 | F | Aggressive |
The weighting is the author's framework and is described in each agent's prompt; it is a way to rank provisions for review, not a legal opinion.
What to expect
- A full review of a long LPA runs all five agents and typically takes 15 to 40 minutes. Shorter documents are faster.
- The report,
LPA-REVIEW-[fund]-[date].md, is written once at the end with the final score. A draft with placeholder scores means the run was interrupted. - On a host without subagents, the same five prompts run inline, in the same order.
- The agents are only used by
review-lpa; the other commands are single skills.