The Rule 11 plea agreement framework
Fed. R. Crim. P. 11 governs every federal plea agreement. Subsection (c)(1) defines three plea types — charge-bargain (A), sentence-recommendation (B), and binding (C) — each with different consequences for the defendant's ability to withdraw if the judge does not follow the bargain.
- Rule 11(c)(1)(A) — charge concession
- The government agrees not to bring, or to dismiss, particular charges. The court is not bound by the concession itself, but acceptance of the plea ordinarily results in dismissal of the bargained-away counts at the sentencing hearing. The defendant cannot withdraw the plea if the court accepts a charge concession and then imposes a high-end sentence on the remaining counts — the only mechanism the (A) plea controls is the indictment, not the punishment. United States v. Hyde, 520 U.S. 670 (1997), confirms that once a court conditionally accepts the plea, it is bound to allow withdrawal only on the limited Rule 11(d) grounds.
- Rule 11(c)(1)(B) — non-binding sentence recommendation
- The parties agree to a recommendation or stipulation regarding sentence — typically a § 3553(a) variance, an offense-level computation, a Chapter 4 criminal-history position, or a specific custodial range. The recommendation is non-binding; the judge may impose any lawful sentence, and the defendant has no right to withdraw the plea if the court declines to follow it. Rule 11(c)(3)(B) requires the court to confirm at the plea hearing that the defendant understands the recommendation is not enforceable against the court. The (B) plea is the workhorse in N.D./E.D. Texas — most fraud, drug, and weapons-offense pleas are structured this way.
- Rule 11(c)(1)(C) — binding sentence agreement
- The parties agree to a specific sentence, sentencing range, or sentencing factor that — if the court accepts the agreement — binds the sentencing judgment. Rule 11(c)(5) gives the court three options: accept, reject (in which case the defendant must be permitted to withdraw), or defer the decision pending the presentence report. Some district judges in N.D./E.D. Texas categorically refuse to entertain (C) pleas; others accept them in specific contexts (cooperator agreements, fast-resolution white-collar cases). Freeman v. United States, 564 U.S. 522 (2011), and Hughes v. United States, 138 S. Ct. 1765 (2018), govern the eligibility of (C) plea defendants for § 3582(c)(2) reduction.
- The Rule 11(b) colloquy
- Before accepting any plea, the court must conduct a colloquy personally addressing the defendant in open court — confirming voluntariness, knowledge of the rights waived, the maximum penalty, any mandatory minimum, the existence and terms of any appeal waiver, and the factual basis under Rule 11(b)(3). The plea hearing in N.D./E.D. Texas typically runs 30 to 60 minutes per defendant; magistrate judges may conduct plea colloquies on referral under United States v. Williams, 23 F.3d 629 (2d Cir. 1994), with the district judge taking the final acceptance. Counsel's job during colloquy is to ensure the record is clean of any ambiguity that could later support a withdrawal or § 2255 attack.
The structural simplicity of Rule 11 — three plea types, one colloquy framework, one withdrawal standard — disguises the analytical depth of any actual plea negotiation. Every paragraph of the written plea agreement reshapes the defendant's sentencing exposure: the offense-level stipulations control the starting Guideline calculation, the criminal-history acknowledgments control Chapter 4 scoring, the appeal-waiver scope controls post-judgment relief, the cooperation language controls the government's discretion to move for a § 5K1.1 departure, and the relevant-conduct stipulations control the upward-adjustment exposure under U.S.S.G. § 1B1.3. Plea-agreement counsel reads each clause as a sentencing input, not as a separate term — and refuses to let any one paragraph fix exposure before the rest of the agreement is on the table.
In N.D. Texas (Dallas, Fort Worth, Lubbock, Amarillo, San Angelo, Abilene, Wichita Falls divisions) and E.D. Texas (Sherman, Tyler, Marshall, Beaumont, Texarkana, Lufkin divisions), plea practice varies by judge. Some prefer detailed factual recitations on the record; some accept written factual bases without extensive oral colloquy; some refuse to conditionally accept the plea and defer acceptance until the presentence report is filed under Rule 32. Counsel who appears regularly in these courts adjusts the written agreement to the particular judge — including the depth of the factual basis, the presence or absence of a (C) binding-sentence clause, and the timing of the cooperation agreement (separate addendum vs. integrated paragraph).
Choosing among (A), (B), and (C) plea types
The (A) charge-bargain controls the indictment but not the sentence. The (B) sentence-recommendation gives the parties a unified position but no enforceable cap. The (C) binding plea locks the sentence — at the cost of judicial acceptance risk. The right choice depends on the judge, the Guideline range, and the cooperation posture.
The choice among Rule 11(c)(1)(A), (B), and (C) is the single most consequential structural decision in any federal plea negotiation. Each carries different risks and rewards, and each interacts differently with the post-judgment landscape — § 3582(c)(2) sentence reduction, § 2255 collateral attack, supervised-release modification, and compassionate release under § 3582(c)(1)(A)(i). Plea-agreement counsel runs the math on each option before recommending one — projecting the likely Guideline range, the realistic § 3553(a) variance posture, the appeal-waiver scope, and the cooperation-departure expected value.
A (A) charge-bargain is appropriate where the indictment is overcharged, where some counts carry mandatory minimums that the bargained-away counts also carry, or where the defendant's primary interest is in avoiding a specific stigmatic charge (sex-offender registration, deportation predicate, professional-license forfeiture). The (A) bargain does not control the sentence — but it can dramatically reduce the offense-level computation under U.S.S.G. § 1B1.2 by removing counts that drive the grouping analysis. In a multi-count indictment with overlapping grouping under § 3D1.2, a (A) bargain that drops two of five counts may have no Guideline effect at all; in a non-grouping case it can eliminate years of exposure.
A (B) sentence-recommendation is appropriate where the parties agree on the Guideline computation, the appropriate variance posture, and the cooperation framing — but where the judge's sentencing pattern is not predictable enough to risk a (C) rejection. The (B) plea gives the parties a unified position at sentencing, a fully briefed § 3553(a) variance request, and a clean record for appellate review of any sentence outside the agreed range — without the catastrophic risk that the judge rejects the agreement and forces the defendant either to withdraw (losing the bargain and exposing every count) or proceed to sentencing with the bargain in disarray. Most N.D./E.D. Texas pleas are (B) pleas for precisely this reason.
A (C) binding-sentence plea is appropriate only where (i) the judge is known to accept (C) pleas in the relevant offense category, (ii) the bargained sentence is materially below the realistic projection under a (B) plea, and (iii) the defendant has assessed the rejection risk. If the judge rejects a (C) plea under Rule 11(c)(5), the defendant must be permitted to withdraw — but the withdrawal returns the defendant to the original indictment with the cooperation history (if any) already disclosed to the prosecution. The asymmetry favors the government: it has heard the defendant's cooperation pitch, it has the factual basis filed of record, and the defendant cannot un-ring those bells if the (C) is rejected and trial follows. The (C) plea should be reserved for cases where the upside justifies that risk.
Sentencing exposure analysis before the plea
The plea agreement is a sentencing document. Counsel projects the Guideline range — offense level, criminal-history category, applicable enhancements, applicable reductions — before signing any plea papers, and refuses to commit to a plea structure that locks in a worse range than the realistic post-trial projection.
A federal plea agreement is a sentencing instrument first and a guilty-plea instrument second. Every line of the agreement either fixes a sentencing input (offense-level computation, relevant-conduct scope, criminal-history concession, acceptance-of-responsibility credit, cooperation framing) or reshapes the procedural posture of the sentencing hearing (appeal-waiver scope, § 2255 waiver, allocution stipulations). Counsel who treats the plea papers as a generic admission document — without running the Guideline projection — leaves years of exposure on the table.
The Guideline projection begins with the offense-level computation under U.S.S.G. Chapter 2. For a § 841(b)(1)(A) drug case, that means projecting the drug quantity (USSG § 2D1.1 Drug Quantity Table), the role-in-the-offense adjustment (§§ 3B1.1, 3B1.2), the safety-valve eligibility under 18 U.S.C. § 3553(f) and U.S.S.G. § 5C1.2, and any specific-offense-characteristic enhancements (firearm under § 2D1.1(b)(1), maintaining premises under (b)(12), violence under (b)(2)). For a § 1343 wire-fraud case, the loss-amount calculation under § 2B1.1(b)(1) frequently drives the range — a $1M loss adds 14 levels; a $9.5M loss adds 18; the difference can be a decade of custody. Plea-agreement counsel runs both the government's likely projection and the defense projection and identifies every contested input.
Criminal-history scoring under Chapter 4 is the second axis. Status points under § 4A1.1(d) (post-Amendment 821), career-offender designation under § 4B1.1, and Armed Career Criminal Act enhancements under 18 U.S.C. § 924(e) can transform the range entirely. The categorical approach under Mathis v. United States, 579 U.S. 500 (2016), Borden v. United States, 593 U.S. 420 (2021), and Wooden v. United States, 595 U.S. 360 (2022), routinely defeats career-offender designations when the prior convictions are scrutinized — but only if counsel raises the issue before the plea is signed. A plea that stipulates to career-offender status forfeits the categorical-approach argument; a plea that preserves the issue keeps a decade-plus reduction on the table.
Acceptance-of-responsibility credit under U.S.S.G. § 3E1.1 is the third structural input. The 2-level reduction under § 3E1.1(a) is available on plea; the additional 1-level reduction under § 3E1.1(b) requires a government motion and turns on timeliness — the plea must be entered early enough to permit the government to allocate its resources. Counsel structures the plea to qualify for the full 3-level reduction and to preserve any argument for the § 3E1.1(b) third-level motion in the written agreement. A plea entered the morning of trial typically does not qualify for the third level; a plea entered at the first pretrial conference does.
The realistic post-trial Guideline range is the floor for plea negotiation. If the projected post-trial range is 78-97 months (Level 28, Category I) and the plea offer produces 87-108 months (Level 29, Category I after losing acceptance-of-responsibility), the plea is a worse outcome and should not be accepted. Counsel runs this math openly with the client and rejects pleas that do not produce a Guideline-range or variance advantage. The advantage may come from acceptance-of-responsibility (3 levels), safety-valve eligibility (2 levels plus relief from the mandatory minimum), cooperation departure projection (variable but historically 30-50%), or charge-bargain count elimination — but it must be quantified before signing.
Cooperation framing — 5K1.1 and the § 3553(e) parallel
A 5K1.1 substantial-assistance motion under U.S.S.G. § 5K1.1 reduces the advisory Guideline range; a parallel motion under 18 U.S.C. § 3553(e) is required to break a statutory mandatory minimum. Both lie in sole prosecutorial discretion — and both are negotiated, not granted.
Cooperation under U.S.S.G. § 5K1.1 is the single largest reliable downward departure available in federal sentencing. The Sentencing Commission reports a median reduction of approximately 47% from the otherwise-applicable Guideline minimum on a § 5K1.1 grant — though the actual reduction varies widely with the quality, timing, and provability of the assistance rendered. A § 5K1.1 motion is filed by the government, not by the defense; the defendant cannot compel the motion, and the government's decision is reviewable only on a showing of unconstitutional motive under Wade v. United States, 504 U.S. 181 (1992), and Melendez v. United States, 518 U.S. 120 (1996).
The § 3553(e) companion motion is structurally critical and routinely overlooked. A § 5K1.1 motion alone reduces the Guideline range but cannot break a statutory mandatory minimum. To impose a sentence below a mandatory minimum on the basis of cooperation, the government must file a separate motion under 18 U.S.C. § 3553(e). The plea agreement must obligate the government to file both motions if it intends to honor a cooperation commitment that would otherwise be capped at the mandatory minimum. A plea that recites only a § 5K1.1 commitment leaves the defendant facing the floor — a mistake worth years of avoidable custody.
Cooperation framing in the plea agreement governs timing, scope, and the standard for evaluation. The agreement should specify whether assistance must be "substantial" (the statutory standard) or "useful" (a softer formulation), whether the assistance includes testimony or only debriefing, whether the agreement contemplates a Rule 35(b) post-sentencing motion in the event cooperation continues, and whether the government retains unilateral discretion to evaluate the assistance or commits to good-faith evaluation. The Northern and Eastern District of Texas U.S. Attorney's Offices use template cooperation language; modifications are negotiable but require persistence from counsel.
The Queen-for-a-Day proffer letter is the typical entry point. Before cooperation is memorialized in the plea agreement, defense counsel and the prosecutor exchange a proffer letter that defines the use the government may make of any statements the defendant gives during the proffer session. The standard letter permits the government to use proffer statements for impeachment if the defendant testifies inconsistently at trial, and to use the statements to develop additional evidence — but bars direct use in the case-in-chief. United States v. Mezzanatto, 513 U.S. 196 (1995), upheld waiver of Federal Rule of Evidence 410 protections through a knowing and voluntary proffer agreement. Counsel reviews the proffer letter line-by-line before any debrief and refuses ambiguous use-protection terms.
Rule 35(b) post-sentencing reduction provides a second cooperation pathway. Fed. R. Crim. P. 35(b)(1) authorizes the government to move within one year of sentencing for a sentence reduction based on substantial assistance; Rule 35(b)(2) extends the window beyond one year for cooperation that ripens after sentencing or that the defendant could not earlier have provided. Plea agreements should anticipate Rule 35(b) where the cooperation timeline runs past sentencing — a debriefing schedule, a target-evaluation provision, and a written commitment by the government to consider Rule 35(b) on a defined evaluation date. Melendez v. United States, 518 U.S. 120 (1996), confirms that Rule 35(b) and § 5K1.1 stand on the same footing for purposes of mandatory-minimum relief — both require a § 3553(e) companion to break the floor.
Plea withdrawal under Rule 11(d)
Withdrawal before the court accepts the plea is available for any reason. Withdrawal after acceptance but before sentencing requires a "fair and just reason" under Rule 11(d)(2)(B). Post-sentence withdrawal is barred — relief lies only on direct appeal or § 2255.
Fed. R. Crim. P. 11(d) provides three windows for plea withdrawal, each governed by a different standard. Before the court accepts the plea, the defendant may withdraw for any reason or for no reason — Rule 11(d)(1). After acceptance but before sentence is imposed, the defendant may withdraw only on a showing of "fair and just reason" — Rule 11(d)(2)(B). After sentence, no withdrawal is available; the defendant's only relief is on direct appeal (typically through the appeal-waiver exceptions for ineffective assistance or sentence above the statutory maximum) or on a collateral attack under 28 U.S.C. § 2255.
The "fair and just reason" standard under Rule 11(d)(2)(B) is the most-litigated plea-withdrawal question in the federal courts. The Fifth Circuit applies a multi-factor test under United States v. Carr, 740 F.2d 339 (5th Cir. 1984), and its progeny: (1) whether the defendant has asserted innocence; (2) the prejudice to the government from withdrawal; (3) the delay between the plea and the withdrawal motion; (4) whether withdrawal would inconvenience the court or waste judicial resources; (5) the closeness of assistance of counsel; (6) whether the plea was knowing and voluntary; (7) whether withdrawal would prejudice the government's ability to prosecute. The defendant's subjective change of mind is not, by itself, a fair and just reason; some plausible articulable basis must be presented.
Conditional pleas under Fed. R. Crim. P. 11(a)(2) provide an alternative pathway. A defendant who wishes to preserve a pretrial-motion ruling for appellate review (most commonly a suppression denial under the Fourth Amendment) may enter a conditional guilty plea reserving the right to appeal the specified pretrial ruling — and may withdraw the plea if the appellate court reverses. The conditional plea requires the consent of the court and the government and is comparatively rare in N.D./E.D. Texas practice. United States v. Markling, 7 F.3d 1309 (7th Cir. 1993), articulates the framework most circuits have adopted.
Post-sentence relief from a plea is functionally a § 2255 question. The waivers in most modern plea agreements bar direct appeal of the conviction or sentence except for narrow carve-outs (ineffective assistance, prosecutorial misconduct, sentence above the statutory maximum, sentence above an agreed cap). On a § 2255 motion, the defendant typically alleges that the plea was not knowing and voluntary because counsel failed to advise of a material consequence, or that counsel's deficient advice — about deportation under Padilla v. Kentucky, 559 U.S. 356 (2010), about sentencing exposure, or about an unraised affirmative defense — caused the defendant to plead when a competent defendant would not have. The Strickland prejudice prong in the plea context requires a reasonable probability that the defendant would have insisted on trial; Lee v. United States, 137 S. Ct. 1958 (2017), refined the analysis when deportation was at issue.
Constitutional waivers — appeal, § 2255, and the limits
Modern federal plea agreements in N.D./E.D. Texas typically include both an appeal waiver and a § 2255 collateral-attack waiver. The Fifth Circuit enforces both — with narrow carve-outs for ineffective assistance, prosecutorial misconduct, sentence above the statutory maximum, and the involuntariness of the waiver itself.
Modern plea agreements in the Northern and Eastern Districts of Texas routinely contain three waiver clauses: an appeal waiver under United States v. Bond, 414 F.3d 542 (5th Cir. 2005), a § 2255 collateral-attack waiver under United States v. White, 307 F.3d 336 (5th Cir. 2002), and a Freedom of Information Act waiver. The first two are routinely enforced. The Fifth Circuit's framework for both requires that the waiver be (1) knowing and voluntary and (2) within the scope of the agreement — i.e., the sentence imposed must fall within the parameters the waiver contemplated. United States v. McKinney, 406 F.3d 744 (5th Cir. 2005), confirms that appeal waivers are enforceable even when they bar review of significant sentencing errors.
The carve-outs that survive even an enforced waiver are constitutionally compelled. Under United States v. Hollins, 97 F. App'x 477 (5th Cir. 2004), and the broader Fifth Circuit framework, a waiver does not bar review of (1) a sentence above the statutory maximum; (2) a sentence based on a constitutionally impermissible factor (race, religion); (3) ineffective assistance of counsel claims that go to the negotiation of the plea itself; or (4) a claim that the waiver itself was not knowing and voluntary. The defense routinely preserves these carve-outs explicitly in the written agreement — striking template language that purports to waive all relief and replacing it with the constitutional-floor formulation.
The § 2255 waiver is the more consequential of the two for most defendants. Direct appeal challenges to a sentence within the Guideline range or within an agreed variance cap are rarely successful; the § 2255 motion, by contrast, is the vehicle for the post-conviction claims that most often produce relief — ineffective assistance for failure to advise about a deportation consequence, failure to investigate an affirmative defense, failure to file a notice of appeal when requested, or failure to challenge a categorical-approach issue at sentencing. Roe v. Flores-Ortega, 528 U.S. 470 (2000), established the presumption of prejudice when counsel fails to file a notice of appeal after the defendant requests one; the Fifth Circuit has confirmed that a § 2255 waiver does not bar this claim.
Waiver of the Padilla advisement is impermissible. Padilla v. Kentucky, 559 U.S. 356 (2010), held that counsel has a Sixth Amendment duty to advise a non-citizen defendant of the deportation consequences of a plea. A plea agreement that purports to waive any claim arising from defective Padilla advisement runs into the constitutional floor — the waiver is not enforceable against an ineffective-assistance claim that goes to the validity of the plea itself. The defense ensures the Padilla advisement is on the record at the plea hearing and that the written plea agreement does not contain language that prosecutors or future § 2255 panels might construe as a Padilla waiver.
Fast-track, safety-valve , and § 3553(a) variance posture
Even where formal fast-track is unavailable, the same conceptual structure — early disposition in exchange for sentencing concessions — drives most plea negotiations in N.D./E.D. Texas. Safety-valve eligibility under 18 U.S.C. § 3553(f) and § 3553(a) variance posture are the two most reliable non-cooperation levers.
Formal fast-track programs under U.S.S.G. § 5K3.1 operate primarily in the high-volume border districts — Southern District of Texas, Western District of Texas, District of Arizona, District of New Mexico, Southern District of California — where caseload pressures historically supported a standardized early-disposition departure. The Northern District of Texas and Eastern District of Texas have not adopted general fast-track programs, but case-by-case early-disposition arrangements remain available through prosecutorial discretion and § 3553(a) variance briefing.
Safety-valve eligibility under 18 U.S.C. § 3553(f) and U.S.S.G. § 5C1.2 is the most reliable non-cooperation reduction available for drug-trafficking defendants. The five-factor test under § 3553(f)(1)-(5) requires (1) limited criminal history (now restructured under the First Step Act to permit more defendants), (2) no violence or threats in the offense, (3) no death or serious bodily injury, (4) no leadership role, and (5) truthful disclosure to the government before sentencing. The First Step Act of 2018 (Pub. L. 115-391) significantly expanded eligibility under (1) by replacing the "no more than 1 criminal-history point" requirement with the more generous current formulation. Pulsifer v. United States, 144 S. Ct. 718 (2024), recently resolved the statutory-interpretation dispute over the "and" connecting the (f)(1) subparts.
Safety-valve produces two distinct benefits: (a) relief from the statutory mandatory minimum under 18 U.S.C. § 3553(f), and (b) a 2-level reduction under U.S.S.G. § 2D1.1(b)(18) — collectively a double benefit on drug cases. Plea-agreement counsel structures the safety-valve disclosure carefully: the government will not move for the 2-level reduction unless it accepts the disclosure as complete and truthful, so the disclosure must be thorough but not over-inclusive — the defendant must not admit conduct not reasonably attributable under U.S.S.G. § 1B1.3 relevant-conduct principles.
§ 3553(a) variance is the dominant non-Guideline lever after United States v. Booker, 543 U.S. 220 (2005), and Gall v. United States, 552 U.S. 38 (2007). The seven factors at § 3553(a) — nature and circumstances of the offense, history and characteristics of the defendant, need for the sentence imposed, kinds of sentences available, the Guideline range itself, policy statements, and the need to avoid unwarranted disparities — authorize the court to impose a sentence outside the Guideline range. The plea agreement may stipulate to specific variance arguments, may permit a non-opposition stipulation by the government, or may preserve the defense's right to argue for variance without government opposition. The variance brief is the most important sentencing document the defense files; the plea agreement should preserve maximum room to make that argument.
The First Step Act of 2018 also expanded retroactive relief for crack-cocaine offenses under § 404, modified the stacking of § 924(c) penalties prospectively, and adjusted the good-time credit calculation for inmates under 18 U.S.C. § 3624(b). Plea-agreement counsel evaluates First Step Act eligibility before sentence — both for the current case and for any prior conviction that might support a § 404 motion. A defendant entering a plea today should not be locked out of subsequent retroactive relief by waiver language; the appeal waiver should carve out future statutory or guideline amendments that retroactively reduce the range.
DFW practice considerations in N.D. and E.D. Texas
N.D. Texas (Dallas, Fort Worth, Lubbock, Amarillo, San Angelo, Abilene, Wichita Falls) and E.D. Texas (Sherman, Tyler, Marshall, Beaumont, Texarkana, Lufkin) each have idiosyncratic plea practices — judges who reject (C) pleas, magistrates who run plea colloquies, and prosecutorial offices with template clauses that vary in their willingness to negotiate.
The Northern District of Texas houses seven divisions: Dallas, Fort Worth, Lubbock, Amarillo, San Angelo, Abilene, and Wichita Falls. The Dallas division (Earle Cabell Federal Building) carries the largest plea volume — federal drug, fraud, weapons, and immigration cases originating from Collin, Dallas, Denton, Kaufman, Rockwall, and Ellis counties. Fort Worth (Burnett Plaza) handles cases from Tarrant, Parker, Hood, Johnson, and adjoining counties. Plea practice differs notably across divisions: Dallas magistrates routinely run plea colloquies on referral; Fort Worth judges historically prefer to take all pleas personally. Counsel adjusts the timing and structure of plea papers accordingly.
The Eastern District of Texas houses six divisions: Sherman, Tyler, Marshall, Beaumont, Texarkana, and Lufkin. The Sherman division (Plano and Sherman courthouses) is the highest-volume division for E.D. Texas and handles federal cases originating from Collin, Grayson, Fannin, Cooke, Hunt, and adjoining counties — substantial overlap with the L&L primary service area. Tyler and Marshall divisions are the principal patent-litigation venues for E.D. Texas but also handle a significant criminal-case docket. Plea practice in E.D. Texas tends to be more formalistic than N.D. Texas — written plea papers are favored, factual bases are detailed on the record, and judges take a more active role in the Rule 11 colloquy.
The U.S. Attorney's Offices for both districts maintain template plea agreements that vary in negotiability. The N.D. Texas USAO permits more variance-friendly clauses on white-collar pleas, particularly in the Dallas division — where the bar negotiates regularly with the office and template modifications are routinely accepted. The E.D. Texas USAO is somewhat more conservative on template modification but accepts substantive negotiation on cooperation language, safety-valve framing, and variance stipulations where the underlying case supports them. Counsel who appears repeatedly in these courts develops working relationships with the line AUSAs that ease the negotiation of nonstandard clauses.
The Fifth Circuit's approach to appeal-waiver enforcement is uniformly strict but bounded by the constitutional carve-outs identified above. Counsel preserves carve-outs explicitly: ineffective-assistance claims tied to the plea negotiation itself, sentence above the statutory maximum, sentence based on impermissible factors, and unwaivable § 2255 claims for failure to file a notice of appeal. The Fifth Circuit's decisions in United States v. Hollins and United States v. McKinney establish the floor; the defense should not accept template language that purports to waive all relief.
Local rules and standing orders also affect plea practice. Each division publishes a local-rules supplement and an electronic-filing handbook; some judges have standing orders on plea hearings (factual basis filed in advance, video-arraignment availability, plea calendar scheduling). Counsel reviews the local rules and the assigned judge's standing orders before drafting the written plea agreement to avoid form-over-substance problems at the hearing. The most common avoidable error is a factual basis insufficient to support an enhancement that the government later seeks at sentencing — counsel ensures the factual basis includes every element of every enhancement contemplated by the agreement.
