Bond covenant analysis with agents that quote every basket, price it in dollars and show how value could leave — before books close
Every basket in dollars and every leakage route found, before books close.
See one case, screen by screen ↓Bond covenant analysis, or indenture analysis, is reading a high yield indenture or leveraged loan agreement to work out what the issuer may do to bondholders: how much it can pay out, invest in unrestricted subsidiaries, borrow, secure or sell, and whether baskets can be chained to move value away. Bond Covenant Analysis gives the credit analyst seven agents that quote every basket with its page, price it in dollars, score it against the firm's own library and red-team the document — and the analyst submits the score for a senior analyst to approve.
AI in asset management: a buyer's guide to agentic AI for investment managers →
Updated
A high yield bond or leveraged loan is protected only by the words in its indenture or credit agreement: how much the issuer may pay out under the restricted payments covenant, how much it may invest in unrestricted subsidiaries the bondholders cannot reach, how much debt and how many liens it may add, and what it must do with asset-sale cash. Each limit is a basket, and the modern ones grow with earnings — “the greater of $50 million and 15% of EBITDA”. While a deal is marketed, the credit analyst has one to three days to read a 300 to 450-page offering memorandum and every defined term those covenants lean on, turn each grower and builder basket into dollars in Excel, compare the covenant package with recent deals and write the covenant section of the memo. A 2021 review of covenant scoring put it plainly: covenant analysis “is very manual in nature, and requires specialist subject matter expertise”.
The risk sits in the route nobody had time to chain. A builder that counts net income from before the issue date, a general basket that can also fund investments, an unrestricted-subsidiary basket with no cap: each looks modest alone, and together they can move a company's most valuable assets out of the bondholders' reach; the market's J.Crew blocker exists because trademarks once left that way. Liability management exercises are now routine, most leveraged loans are covenant-lite, the best-known covenant service covers only deals of $250 million and up, and the held book is rarely re-read when EBITDA moves or an amendment loosens a document.
Estimated split for one high yield new issue with a 400-page offering memorandum, by hand and with the solution.
Seven agents quote, price, rank and attack every basket; an analyst submits, a senior analyst signs.
Dunmore Cinemas Holdings, a theater chain owned by the sponsor Halberd Ridge Partners, is marketing $650 million of 7.25% senior notes due 2031, offered under Rule 144A and Reg S, to refinance notes Quillfield Capital already holds. Books close Thursday at 12:00 p.m. The 412-page preliminary offering memorandum arrived Tuesday at 6:05 p.m. Here is the high yield covenant review that followed, screen by screen, in the working solution.
Nadia covers 46 issuers, and six new issues, $3.4B in all, are in market this week. Her Today screen opens on what needs her: Dunmore scored 2.1 with $212.0M that can leave, Pellston Paper using a route flagged in March, Valdemar Telecom scored from its Spanish half-year report, and Corrigan Health held because pages 311–314 are unreadable. Beside it, what the agents did since yesterday, every line cited to its page, and a box to ask them anything.
“Read Dunmore’s 412-page offering memorandum, scored it 2.1 of 5 and found a route for $212.0M to leave.”
The Reader quoted 41 covenant provisions and 63 defined terms, each with section and page. The scorecard opens on a bank of five basket capacity meters from LTM EBITDA of $386.0M: the restricted payments builder at $79.1M, the general restricted payments basket at $57.9M, general liens at $100.0M, asset sales at $60.0M and investments in unrestricted subsidiaries at $75.0M — looser than 88% of peers. Beneath them, seven blocker badges: J.Crew, unrestricted cap and Chewy missing; Serta present.
Reader agent: “One thing to look at first: the builder counts net income from January 1, 2022, almost five years before the issue date.”
Nadia opens the general basket. LTM adjusted EBITDA of $386.0M for the twelve months to June 30, 2026, times 15%, is $57.9M against a $50.0M floor; nothing used to date; and because “Restricted Payment” includes a Restricted Investment, the full amount counts as investment capacity. Each line links to its page — 31, 87 or 49.
“The Verifier agent recomputed every line independently and got the same result.”
The peer set is built from Quillfield's own library of 2,860 bonds and loans: US high yield, B-rated, sponsor-backed, consumer and leisure, issued 2024–2026 — 318 deals. On the unrestricted basket Dunmore sits at 19.4% of EBITDA against a 10.4% median. The nearest comparable deals without a J.Crew blocker score 1.7 to 2.3.
The Adversary works the document the way a sponsor's counsel would. Step one puts the trademarks and the LoyaltyPass platform into a new restricted subsidiary; step two designates it unrestricted; step three funds the deemed investment from three baskets — $75.0M, $57.9M and $79.1M, chained to $212.0M against the IP's $171.4M carrying value. Eleven routes tested: one open, one to watch, nine blocked. The next morning Nadia asks why, and the answer cites pages 62, 101 and 214.
“Nothing in the definitions, §4.07 or §4.15 limits material IP: the Reader searched 23 pages.”
A second agent re-read every quote and matched 40 of 41. On the 41st — whether returns on investments reload the $75.0M basket above its original size — the readings differed, so the score could not be published until legal ruled. Helena ruled for the Verifier's reading: returns restore the basket up to $75.0M; they do not enlarge it. The same screen holds an open dispute on Maple Harbor, with reading A and reading B side by side.
The Verifier never resolves its own dispute; disputed readings block publication.
Nadia reviews the drafted recommendation — do not participate at current terms; participate at up to $45.0M if the issuer adds a J.Crew blocker and caps investments in unrestricted subsidiaries at $40.0M — and submits at 9:05. Graham Whitlow, senior credit analyst, approves at 9:31. Because the score is below 2.5, Catherine Ashby, Head of Credit Research, gives the second approval at 9:48. The score reaches the high yield PMs, and a pre-trade note goes to the order management system for three funds.
“Published to Tobias Lund and the high yield PMs. Pre-trade note added in the order management system for 3 funds.”
The Memo agent drafted four asks: a J.Crew blocker, which closes the IP route; a $40.0M cap on investments in unrestricted subsidiaries; the builder from the issue date, not January 1, 2022; and a Chewy blocker. Together they cut the capacity to move other assets from $212.0M to $137.9M. Nadia submits them, Graham signs, and they go to Harlan Street Securities' leveraged finance syndicate. The agents will compare the final offering memorandum with the asks.
“Nothing is sent without a senior analyst’s signature.”
Across 418 issuers and 1,146 positions, $1.31B could leave held bonds. Pellston Paper designated Pellston Specialty Fibers unrestricted yesterday in an 8-K — the route the Adversary flagged on March 12 — and its score fell from 2.3 to 1.6 with $48.6M held in three funds. One click lists the 14 held deals with Dunmore's gap: no J.Crew blocker and an unrestricted basket over 15% of EBITDA, $1.71B of positions.
Brightwater Pet Brands' term loan Amendment No. 3 lifts the general liens basket from the greater of $100M and 50% of EBITDA to the greater of $250M and 100% of EBITDA, so $118.5M of new debt could rank ahead of the existing term loan. The Amendment agent redlined it against the current document and re-scored it from 2.9 to 2.4; the Memo agent recommends do not consent, despite a 25 bp fee on $67.0M held across the loan fund and three CLOs.
“Your instruction needs a second person before it goes to the custodians.”
The same new issue, seen by five of the people who carry it — what the books window looked like, and what it looks like now.
Seven agents quote, price, rank and attack every basket; an analyst submits, a senior analyst signs.
Reads indentures, offering memoranda, credit agreements, supplements and amendments in any language; quotes every covenant, basket and defined term with its section and page, and follows the cross-references.
Turns grower and builder baskets into dollars from the latest twelve-month EBITDA, total assets and net income, from the offering memorandum or the issuer's own filing in its language, and shows its working.
Ranks every basket against the peer set in the firm's own covenant library — rating, region, sponsor, sector, vintage — and scores the deal on the house rubric.
Plans and chains routes for value to leave the way a sponsor's counsel would — drop-down, designation, priming, guarantor release, uptier — prices each in dollars and lists what would stop it.
Re-reads every quote and recomputes every dollar independently, with a second model; tries to break each route and raises disputes to legal counsel with both readings.
Watches amendments, consent solicitations, waivers and 8-K designations for held deals; redlines each against the current document, re-scores it and dates the deadline.
Drafts the covenant section of the credit memo, the asks for the syndicate desk ranked by dollars of leakage closed, and the monthly covenant report from approved scores; tracks the final terms against the asks.
Maker, checker, rulings, consents. The agents propose; a named person decides.
Every working screen keeps the agents' conversation in place. Ask in plain words or give a what-if; answers cite their pages, and a what-if never changes the score of record.
Why does the general basket count?
Because clause (15) allows any “Restricted Payment” up to the greater of $50.0M and 15% of EBITDA (§4.07(b)(15), p. 87), and the definition of “Restricted Payment” includes any Restricted Investment (p. 49). An investment in an unrestricted subsidiary is a Restricted Investment, so the full $57.9M can fund one. The Verifier read it the same way.
What if EBITDA falls 20%?
At $308.8M of EBITDA the floors take over: the general basket drops to its $50.0M floor, the builder starter stays at $40.0M and the $75.0M basket is fixed. The route still carries $204.1M — the floors keep 96% of today's capacity, so a weaker year does not close the route.
Cap unrestricted at $40M, re-score
Done as a what-if, not saved: with clause (18) at $40.0M the route carries $177.0M and the score rises to 2.3. The IP ($171.4M) still fits, so only a J.Crew blocker closes the IP route. With both changes the score would be 2.9, second quartile.
Which routes hang on one clause?
Two. The uptier is blocked only by the Serta blocker (§9.02(a)(9), p. 124) and layering only by the anti-layering covenant (§4.11, p. 95). If an amendment removed either, that route would open. The Amendment agent watches both clauses in every consent solicitation for this deal.
13 screens from the working solution, on its sample data. Pick one to see it large.
The thresholds every score and instruction passes through are settings. Changes that touch scoring go for the Head of Credit Research's approval and onto the audit trail.
| Setting | Default | Choose from |
|---|---|---|
| Second approver when the score is below | 2.5 | Head of Credit Research approves after the senior analyst |
| Second approver when a route exceeds | $150.0M | Head of Credit Research approves after the senior analyst |
| Agent confidence below which a person must check | 0.90 | Applies to every agent |
| Asks to a syndicate desk need | 1 senior analyst | Maker-checker; the analyst submits, never sends |
| Consent instructions need | PM + legal counsel | Maker-checker before the custodians and agent banks receive it |
| Disputed readings block publication | On | Legal counsel rules |
| Scoring rubric | v3.2 — eight weighted sub-scores | Restricted payments 20% · investments and unrestricted subsidiaries 20% · LME blockers 15% · debt 15% · liens 10% · asset sales 10% · change of control 5% · reporting 5%; roll back to v3.1 or v3.0 |
| New issue scored | 4 hours before books close | Escalates to the senior analyst at 50% of the time left |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate (where a new issue's analyst hours go) · “typical” = published figures (a November 2021 literature review of quantitative approaches to bond covenant scoring — covenant analysis “very manual in nature”; Q4 2024 research on liability management exercises — 91% of US leveraged loans by par covenant-lite; a credit research firm's annual count — 34 liability management exercises in 2024 and 28 in 2025; the published coverage of the best-known covenant service — high yield bonds and syndicated loans of $250 million and up). J.Crew, Chewy and Serta are the market's names for blocker provisions, after public transactions. Quillfield Capital and the people, issuers, sponsors, banks, funds, identifiers and figures named on this page are fictional — characters and sample data in the working solution.
Bond covenant analysis, or indenture analysis, is reading a high yield indenture or leveraged loan agreement to work out what the issuer may do to bondholders: how much it can pay out, invest in unrestricted subsidiaries, borrow, secure or sell, and whether baskets can be chained to move value away. Bond Covenant Analysis gives the credit analyst seven agents that quote every basket with its page, price it in dollars, score it against the firm's own library and red-team the document — and the analyst submits the score for a senior analyst to approve.
A J.Crew blocker stops material intellectual property from being transferred to, or held by, an unrestricted subsidiary; it is named after the transaction that made that route well known. The scorecard shows it beside the Chewy and Serta blockers, anti-layering, the non-guarantor cap and the change of control put as present, partial or missing. When it is missing, the source passage shows where the Reader searched, and the Verifier repeated the search independently.
The Calculator agent takes the latest LTM EBITDA, total assets and net income from the offering memorandum or the issuer's own filing. A grower basket is the greater of its dollar amount and its share of EBITDA; a builder adds a share of cumulative net income from its build date, and a build date set before the issue date is flagged. Every line of the working is cited and recomputed by a second agent.
The Adversary agent plans routes the way a sponsor's counsel would — drop-down, designation, priming, guarantor release, uptier — chains the baskets and definitions, and prices how much value can leave today. The Verifier tries to break each route; only routes both agree on are shown as open, each step quoted, with the asks that would close it ranked by dollars.
The Amendment agent watches amendments, consent solicitations, waivers and 8-K designations, redlines each against the current document, re-scores the deal and dates the deadline. Held bonds are also re-scored when new quarterly numbers arrive. The PM decides on a consent and a second person checks the instruction before it goes to the custodians.
Yes. Covenant quality is benchmarked against your own library — in the working solution, 2,860 bonds and loans including 512 deals under $250 million and 640 private and local-law agreements — with peer sets you choose. The scoring rubric is versioned, owned by the Head of Credit Research, and can be rolled back; every percentile shows the peers behind it.
Yes. The agents never publish, send, instruct or trade. The analyst can edit a reading with a reason and submits; a senior analyst approves; the Head of Credit Research gives a second approval below 2.5 or for a route over $150.0M; legal counsel rules on disputed readings. Every step is on a hash-chained audit trail.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your rubric, a sample of indentures and credit agreements, and your holdings — and it goes live once every quality gate has passed. A guided first-time setup connects the document store, financial data feed and holdings, and imports your library.
Every credit question answered in minutes, each sentence pinned to its tranche and page.
1-min demoProspectus ReviewEvery deal that fits a fund briefed and decided before books close, every line cited.
1-min demoInternal Credit RatingEvery rating cited to its page, and every two-notch split explained before committee.
1-min demoWe'll run Bond Covenant Analysis on a sample of your own indentures, credit agreements and held positions.