Before Smith, Southwick, and Higginson, Circuit Judges. Jerry E. Smith, Circuit Judge: Larry Lincks pleaded guilty, then filed a 28 U.S.C. § 2255 motion seeking relief from his plea. The district court denied the motion, and Lincks appeals. This case reveals uncertainty in our caselaw regarding argument for-feiture, ineffective assistance of counsel (“IAC”) in guilty pleas, and the nature of non-constitutional appeals in § 2255 proceedings. After addressing and resolving those uncertainties, we affirm the denial of the § 2255 motion. We also clarify that COAs are unnecessary for objections to evidentiary rulings in § 2255 proceedings, but on the merits we reject Lincks’s evidentiary challenge. I. Lincks pleaded guilty of possession with intent to distribute a mixture or substance containing a detectable amount of methamphetamine[1] and was sentenced to 188 months of imprisonment, to be served consecutively to any sentence imposed in two unrelated state proceedings, and three years of supervised release. We dismissed his appeal, and the Supreme Court denied certiorari. Lincks v. United States, 139 S. Ct. 2731 (2019). Lincks then filed a § 2255 motion, contending, among other things, that his attorney rendered IAC because (a) he failed to explain the presen-tence report (“PSR”), making the guilty plea unknowing and involuntary; (b) he failed to file a motion to suppress and dismiss evidence and failed to investigate; (c) he erroneously advised that Lincks’s guideline range would be 60 to 72 months[2]; and (d) he failed to look at discovery materials, which led to a higher guideline range than expected. Lincks submitted copies of e-mails between him and his counsel, Aaron Wiley, as exhibits. They revealed that shortly before Lincks signed his plea agreement, he e-mailed Wiley stating that he was concerned because he was facing twenty years of imprisonment and that he “[might] as well” go to trial if he was looking at that much time. Wiley responded that the statutory maximum was twenty years and had to be included in the plea agreement, but Lincks’s sentence would not approach the statutory maximum given the drug quantity attributed to him and his criminal history consisting of two prior drug convictions. About eight months later, following the plea and preparation of the PSR, Wiley stated in another e-mail that the guideline range in the PSR was “higher than expected” because Lincks had received more criminal history points and was held responsible for a higher drug quantity than expected. Counsel stated he would try to get a lower guideline range by making objec-tions and would discuss the available options with Lincks. Lincks then told Wiley that the PSR incorrectly stated he had bought a pound of narcotics from a man named Tony; he denied saying that in his interview and asked counsel to listen to the taped interview. In addition, Lincks filed two motions for discovery, which the magis-trate judge denied. The district court denied Lincks’s renewed motion for discovery, accepted the magistrate judge’s report, and denied a certificate of appealability (“COA”). We, however, granted a COA on the issue of whether Lincks’s trial counsel was ineffective in advising him about the guideline range, affecting his decision to plead guilty. Specifically, we stated, “Lincks has made the required showing as to his claim (1)(b) above, namely, that his trial counsel was ineffective in advising him about his sentencing guidelines range . . . .” But as the order listed Lincks’s claims, claim (1)(b) was Wiley’s “fail[ure] to file a motion to suppress and dismiss evidence and fail[ure] to investigate.” Claim (1)(c) was the guidelines-advice claim. Apparently because of the dis-crepancy, Lincks’s merits briefing focuses almost entirely on his evidentiary and failure-to-investigate claims. We denied a COA on Linck’s other IAC claims and carried with the case his request for a COA on whether the district court abused its discretion by denying discovery. II. In an appeal from the denial of a § 2255 motion, this court reviews the district court’s legal conclusions de novo and its factual findings for clear error. United States v. Cavitt, 550 F.3d 430, 435 (5th Cir. 2008). Claims of IAC are reviewed de novo. See United States v. Scott, 11 F.4th 364, 368 (5th Cir. 2021), cert. denied, 142 S. Ct. 827 (2022). As for discovery, “[t]he dis-trict court’s decision regarding [its] availability . . . is . . . committed to the sound discretion of the district court[] and is reviewed under the abuse of discretion standard.” United States v. Fields, 761 F.3d 443, 478 (5th Cir. 2014) (second ellipsis in original) (quoting Clark v. Johnson, 202 F.3d 760, 765–66 (5th Cir. 2000)). III. There are three issues. First, whether Lincks forfeited his argument that Wiley was ineffective in advising him about his guideline range. The answer is no. Second, assuming that we reach the merits of Lincks’s claim, whether hecan prevail on it. Again, no. And third, whether Lincks is entitled to a COA on whether the district court abused its discretion by denying discovery. We reframe that last question to clarify our circuit’s jurispru-dence and answer it, as the others, in the negative. A. The government argues that Lincks forfeited the only issue for which we granted a COA, i.e., his guidelines-advice claim. Specifically, it points to his opening brief’s conclusory statement that “he was given ‘faulty and erroneous advice’ at the plea stage.” As noted, Lincks’s briefing strategy was apparently the result of a typographical error in this court’s grant of a COA: Our order referred to Lincks’s “claim (1)(b)” rather than to “claim (1)(c).” Given the unusual circumstances of the case, we find no forfeiture. “[P]ro se litigants, like all other parties, must abide by the Federal Rules of Appellate Procedure.” United States v. Wilkes, 20 F.3d 651, 653 (5th Cir. 1994) (per curiam). “[W]hen [a criminal] appellant fails to . . . comply with the rules of the court, the clerk must dismiss the appeal for want of pro-secution.” 5th Cir. R. 42.3.2. Our rules require an appellant to specify his arguments and support them “with citations to the authorities and parts of the record on which the appellant relies.” Fed. R. App. P. 28(a)(8); 5th Cir. R. 28.3(i). “Failure to satisfy the requirements of Rule 28 as to a particular issue ordinarily constitutes abandonment of the issue.” United States v. Miranda, 248 F.3d 434, 443 (5th Cir. 2001) (citing United States v. Beaumont, 972 F.2d 553, 563 (5th Cir. 1992)). On the other hand, “pro se briefs are afforded liberal construction.” McCreary v. Richardson, 738 F.3d 651, 655 (5th Cir. 2013) (per curiam) (citing Haines v. Kerner, 404 U.S. 519, 520–21 (1972) (per curiam)), as revised (Oct. 9, 2013). And “we must liberally construe briefs in determining what issues have been presented for appeal. Miranda, 248 F.3d at 444 (citing SEC v. Recile, 10 F.3d 1093 (5th Cir. 1993)). Applying those legal principles to the facts here, there is no forfeiture. Lincks’s confusion, which appears genuine and persists through his reply brief, stems from our order’s inaccuracy in referring to Lincks’s “claim (1)(b)” rather than the intended “claim (1)(c).” We will not punish Lincks for that, especially remembering that the rules of appellate procedure are meant to “promote the interest of justice.” See 28 U.S.C. § 2073(b). But even if Lincks had forfeited his argument, we would still consider it. “[T]he issues-not-briefed-are-[forfeited] rule is a prudential construct that requires the exercise of discretion.” Miranda, 248 F.3d at 243. That discretion is appropriate here: Any forfeiture is at least partly attributable to this court’s imprecision. We therefore proceed to the merits of Lincks’s guidelines-advice claim.[3] B. Lincks asserts that Wiley provided IAC by erroneously underesti-mating his guideline range, making his guilty plea unknowing or involuntary. To demonstrate IAC, Lincks must show (i) that his counsel’s perfor-mance was deficient because it fell below an objective standard of reasonable-ness and (ii) that the deficient performance prejudiced the defense. Strick-land v. Washington, 466 U.S. 668, 689–94 (1984). A failure to establish either prong defeats the claim. Id. at 697. We resolve the claim against Lincks on the first prong and therefore do not address the second. To demonstrate deficiency, Lincks must show that “counsel made errors so serious that counsel was not functioning as the ‘counsel’ guaranteed the defendant by the Sixth Amendment.” Id. at 687. The alleged error here is inaccurate advice concerning Lincks’s exposure under the sentencing guidelines, which may have vitiated his knowledge of the consequences of the plea. In particular, for a guilty plea to be knowing, the defendant must have a “full understanding of what the plea connotes and of its consequence.” Boykin v. Alabama, 395 U.S. 238, 244 (1969). At a minimum, a guilty plea’s sentencing consequences include the maximum sentence for the offense charged. United States v. Hernandez, 234 F.3d 252, 255–56 (5th Cir. 2000) (per curiam). We digress briefly to note that, at least under our caselaw, the “know- ing” and “voluntary” requirements “embody different concepts.” Id. at 255 n.3. Volition in guilty pleas comes into doubt when the plea is induced by threats, improper coercion, false or empty promises, and the like.[4] On the other hand, as we have just noted, “knowledge” generally implicates a fac-tual understanding of the essential components and consequences of a guilty plea.[5] Under the most favorable reading of his arguments, Lincks alleges that his plea was not “knowing.” Having erected the legal framework and assembled the record evi-dence favorable to Lincks, we find no constitutional deficiency in Wiley’s representation. Wiley informed Lincks via e-mail that “twenty years is the statutory maximum for [his] alleged offense”; “[t]he [statutory maximum language] is included in the plea agreement . . . in order for an accused to make an informed decision about entering a plea of guilty”; and—again— that the statutory maximum was “twenty years.” Admittedly, there are contrary indicia. Wiley noted in an e-mail that Lincks’s sentence would not “approach the statutory maximum” given the amount of narcotics involved and his criminal history.[6] Lincks also asserts that Wiley predicted a sentence between 60 and 72 months; we accept arguendo that factual claim as true. Nevertheless, Wiley provided constitutionally sufficient perfor-mance. In addition to the e-mail representations, Lincks’s plea agreement contained unequivocal and unambiguous statements about his possible sen-tence and Wiley’s representations. Among the most salient in his plea agree-ment were these statements: · “The maximum [applicable] penalties . . . include . . . impris-onment for a period of not more than twenty years . . . .” · “Lincks understands that the sentence in this case will be im-posed by the [c]ourt after consideration of the United States Sentencing Guidelines. The guidelines are not binding on the [c]ourt[] but are advisory only. Lincks has reviewed the guide-lines with his attorney, but understands no one can predict with certainty the outcome of the [c]ourt’s consideration of the guidelines in this case. Lincks understands that he will not be allowed to withdraw his plea if the applicable advisory guide-line range is higher than expected, or if the [c]ourt departs from the applicable advisory guideline range. Lincks will not be al-lowed to withdraw his plea if his sentence is higher than ex-pected. Lincks fully understands the actual sentence imposed (so long as it is within the statutory maximum) is solely left to the discretion of the [c]ourt.” · “There have been no guarantees or promises from anyone as to what sentence the [c]ourt will impose.”[7] · “[The plea agreement] is a complete statement of the parties’ agreement . . . .” Lincks does not argue that Wiley misled or misinformed[8] him about any of those provisions. In summary, we assume arguendo that Wiley suggested a particular and inaccurate guidelines range to Lincks. But Lincks also, in the most ex-press terms possible, disclaimed any reliance on that specific suggestion, acknowledged that he had discussed the guidelines with Wiley and under-stood that the district judge had complete discretion over his sentencing regardless of the guidelines range,[9] denied that Wiley had promised or guar-anteed any particular sentence, and admitted knowledge of the maximum possible sentence, which Wiley indisputably communicated to Lincks over e-mail, even labeling it Lincks’s “exposure.” Wiley’s representation of Lincks was not perfect. But not every attor-ney error is constitutionally fatal: “[T]he Sixth Amendment entitles a crim-inal defendant to reasonable, but not perfect, representation of counsel.” United States v. Valdez, 973 F.3d 396, 404 (5th Cir. 2020). Taking all of Wiley’s representations and counsel together, there is no IAC. See id. (find-ing no constitutionally deficient performance and no prejudice on materially indistinguishable facts).[10] C. Lincks asserts the district court abused its discretion in denying his discovery motions seeking production of certain communications he had with Wiley. We take this opportunity to clarify our circuit’s law on the need vel non for a COA on non-merits claims. Defendants seeking relief under § 2255 generally must obtain a COA. See 28 U.S.C. § 2253. COAs “may issue . . . only if the applicant has made a substantial showing of the denial of a constitutional right.” Id. § 2253(c)(2). But that statutory provision does not address evidentiary and other collateral rulings by the district court—do they require a COA for appealability? Despite uncertainty in our precedent,[11] our answer is no, at least as to evidentiary rulings. Instead, they are appealable as needed to resolve consti-tutional claims subject to COAs under § 2253. In other words, although a defendant need not obtain a COA for an evidentiary ruling, that ruling is sub-ject to direct appellate review only insofar as it relates to a claim for which a COA is sought or has been granted.[12] Thus, “a petition challenging an evi-dentiary ruling may only be entertained as corollary to a constitutional viola-tion.”[13] At a practical level, a § 2255 petitioner seeking appellate review of an evidentiary decision must explain why review of that decision is necessary or helpful in resolving a constitutional claim for which he is seeking (or has obtained) a COA. Treating Lincks’s request for a COA as a direct appeal, the district court did not abuse its discretion. A § 2255 movant is not entitled to discov-ery, but discovery may be authorized for good cause under Rule 6(a) of the Rules Governing § 2255 Proceedings. See Fields, 761 F.3d at 478. A movant “demonstrates ‘good cause’ under Rule 6(a) ‘where specific allegations before the court show reason to believe that the petitioner may, if the facts are fully developed, be able to demonstrate that he is . . . entitled to relief.’” Id. (ellipsis in original) (quoting Bracy v. Gramley, 520 U.S. 899, 908–09 (1997)). The district court acted within its discretion in finding that Lincks failed to meet that standard. As the court noted, his arguments on his guidelines-advice claim are conclusory. And his request for discovery was based on a bare assertion that the requested material “will establish the valid- ity of” his constitutional claim. Moreover, we have now rejected Lincks’s guidelines-advice claim, and there is no indication either in his briefing or in the record (which arguendo we interpret liberally in his favor) that the discov-ery he seeks would conceivably produce a different result. See id. (“We have noted that Rule 6 . . . ‘does not authorize fishing expeditions.’” (Ward v. Whitley, 21 F.3d 1355, 1367 (5th Cir. 1994))). * * * * * The denial of Lincks’s § 2255 motion is AFFIRMED. Lincks also requested a COA as to whether the district court erred in denying his dis-covery requests. In the interests of justice and to ensure conformity with our precedent, we construe that request as a direct appeal and AFFIRM the denial of Lincks’s discovery requests. Stephen A. Higginson, Circuit Judge, concurring: I agree in full with the panel’s analysis regarding Lincks’s preservation of his claim and the appealability of his discovery issue. I also agree that, under our court’s case law, Lincks’s claim for ineffective assistance of counsel (“IAC”) fails on the deficiency prong. I write separately to explain why I have concerns about that case law and its approach to assessing sentencing-advice errors that result in guilty pleas. Perhaps the most important decision a criminal defendant makes is whether to proceed to trial or to plead guilty. It is, therefore, of paramount importance that defense counsel advise their clients on all aspects and consequences of entering a guilty plea. One year after the Supreme Court established its two-part test for assessing IAC claims in Strickland v. Washington, 466 U.S. 668 (1984), the Court held that the same test applies to IAC-based attacks on guilty pleas. Hill v. Lockhart, 474 U.S. 52, 58 (1985). Under that test, a defendant challenging his guilty plea on the ground of IAC “must show that counsel’s representation ‘fell below an objective standard of reasonableness’ and that he was prejudiced as a result.” Lee v. United States, 582 U.S. 357, 363 (2017) (quoting Strickland, 466 U.S. at 688). Here, Larry Lincks contends that his guilty plea is constitutionally infirm because his plea counsel underestimated his prison-sentence exposure by almost ten years. Our panel concludes that Lincks’s claim fails on the deficiency prong, primarily citing our court’s recent decision in United States v. Valdez, 973 F.3d 396 (5th Cir. 2020). In Valdez, after a jury was empaneled for his trial, the defendant pleaded guilty to possessing a firearm as a convicted felon. Id. at 399. But Valdez’s lawyer failed to tell him that his offense level under the Sentencing Guidelines would be determined by a cross-reference to a homicide offense. Id. at 400-02. Counsel estimated that Valdez’s Guidelines range would be between twenty-four and thirty-three months. Id. at 399. But, with the cross-reference, his range was actually 324 to 405 months. Id. at 401. The district court sentenced him to the statutory maximum of ten years. Id. Valdez filed a § 2255 petition seeking to vacate his conviction, alleging that counsel’s error prejudiced him in his decision to enter a guilty plea. Id. at 399. Our court concluded that there was no Strickland deficiency. In relevant part, we explained: [A]lthough Valdez’s counsel’s estimate of what he “hoped” Valdez’s sentence would be was well below the 120 months to which Valdez was ultimately sentenced, counsel properly apprised Valdez, prior to his pleading guilty, of the maximum penalty the court could impose. And counsel also made abundantly clear to Valdez that no estimation he offered was a guarantee or a promise. [There] is no doubt that counsel’s estimated sentencing range was far lower than Valdez’s actual sentence. Nevertheless, we do not find any deficiency in counsel’s estimation to be unreasonable. Id. at 404. As the panel observes, that analysis appears to resolve the instant case. See ante, at 9 (citing Valdez as finding no deficient performance on “materially indistinguishable facts”). Here, Lincks’s lawyer told him that the statutory maximum under his guilty plea was twenty years. And although counsel allegedly estimated that the sentence would be many years lower than that, Lincks, who is pro se, does not contend that this prediction was a “guarantee or promise.” Unable to convincingly distinguish Valdez, I am compelled to conclude that Lincks’s counsel was not Strickland ineffective under circuit law. But I am apprehensive if this is the Strickland bar for plea counsel. Under this doctrine, it would seem1 that counsel advising a client on whether to enter a guilty plea discharges his constitutional duty as it pertains to sentencing exposure so long as he tells his client the statutory maximum and refrains from making any “guarantee[s]” or “ promise[s].” Valdez, 973 F.3d at 404. But as Judge Wiener observed in dissent in Valdez, other federal courts of appeals require plea counsel to avoid making gross mischaracterizations of the Guidelines estimations for their clients. Id. at 407-08 (Wiener, J., dissenting) (citing cases). Unreasonably incorrect Guidelines calculations may well amount to Strickland deficiency, or at least trigger an obligation of further inquiry. See United States v. Brooks, No. 22-30369, __ F.4th __, 2023 WL 5163315, at *5 (5th Cir. Aug. 11, 2023) (noting that the defendant’s postconviction allegations of deficient performance by plea counsel “may warrant factual inquiry” in § 2255 proceedings). Critically, under the permissive Valdez standard, neither the nature, cause, nor magnitude of counsel’s mistake appears to matter.2 In Valdez, counsel failed to notice a Guidelines cross-reference to homicide offenses in a case where the defendant, pleading guilty to unlawful gun possession, used that gun on the night in question to shoot and kill someone outside his home. And counsel’s estimation of the Guidelines range was more than ten times lower than the defendant’s properly calculated range. But neither of these facts appeared to undermine our court’s conclusion that “any deficiency in counsel’s estimation [was not] unreasonable.” Id. at 404. Notably, this permissive deficiency standard in guilty-plea cases stands in contrast to our treatment of similar errors in the context of guilty pleas forgone. In that context, as we acknowledge, our case law recognizes that, when a defendant declines to enter a guilty plea and instead proceeds to trial, counsel performs deficiently by overestimating the sentence that would result from the guilty plea, e.g., United States v. Rivas-Lopez, 678 F.3d 353, 355, 357-59 (5th Cir. 2012), or by underestimating the sentence that would result from a conviction at trial, e.g., United States v. Grammas, 376 F.3d 433, 436-37 (5th Cir. 2004); United States v. Herrera, 412 F.3d 577, 579, 581 (5th Cir. 2005). We have said, quite plainly, that “[b]y grossly underestimating the defendant’s sentencing exposure, counsel breaches his duty as a defense lawyer in a criminal case to advise his client fully on whether a particular plea to a charge appears desirable.” Grammas, 376 F.3d at 436-37 (cleaned up) (citations omitted). Contrastingly, in Valdez, our court held that plea counsel performed adequately even when he grossly underestimated the sentence the defendant would receive from his guilty plea. See 973 F.3d at 404. The apparent logic for why the “go-to-trial” deficiency does not apply with equal force to the “don’t-go-to-trial” scenario may turn on defendants’ Rule 11 Boykin colloquy statements, made under oath, disclaiming reliance on Guidelines “promises,” see Fed. R. Crim. P. 11(b)(2), as well as the reality that defendants can seek to withdraw their guilty pleas after having seen Guidelines calculations in their presentence reports, see id. 11(d). But without adversarial briefing, I am unsure whether either of these features of guilty-plea cases speaks to whether counsel performed deficiently in the first place by grossly underestimating his client’s sentencing range before the plea was entered. Here, Lincks has alleged that counsel advised him that his Guidelines range would be five to six years’ imprisonment. That estimation missed the mark by almost ten years. And while the record lacks direct evidence of counsel’s numerical estimation, correspondence between Lincks and his lawyer tends to corroborate Lincks’s claim. Before entering his guilty plea, Lincks emailed his lawyer to express concern about the twenty-year maximum. He said that most of the men in prison with him who “had a cap” “got the cap.” Counsel responded, also by email: “[G]iven the amount of narcotics attributed to you . . . and your criminal history (two prior drug convictions) — you won’t approach the statutory maximum.” At sentencing, the two reasons counsel had given for his low prediction—drug quantity and criminal history—served not to limit Lincks’s sentencing exposure but instead to increase Lincks’s Guidelines range substantially. Namely, Lincks was held accountable for over 500 grams of methamphetamine, elevating his base offense level to thirty. And his criminal history consisted of far more than “two prior drug convictions.” Lincks received criminal-history points for six prior convictions, including three drug convictions and three convictions relating to hiding fugitives or obstructing justice. Perhaps there were circumstances making counsel’s (mis)advice reasonable. For example, counsel may have reasonably believed that Lincks would not be held accountable for the over 500 grams of methamphetamine that ultimately found its way into the presentence report. Or, for example, there may have been conversations between Lincks and his lawyer where these risks and possibilities were carefully discussed. But given the evidence of error on the limited record we have, but for our court’s case law, I would hold that Lincks is entitled to an evidentiary hearing to allow the district court to confirm whether plea counsel’s underestimation amounted to Strickland deficiency. Of course, to obtain relief under § 2255, deficient performance alone is not enough. Lincks must also show that he was prejudiced by the deficiency. The Supreme Court has instructed that the prejudice inquiry in the guilty-plea context asks whether there is “a reasonable probability that, but for counsel’s errors, [the defendant] would not have pleaded guilty and would have insisted on going to trial.” Lee, 582 U.S. at 364-65 (quoting Hill, 474 U.S. at 59). Here, too, the record before us warrants further factual development. Before his guilty plea, Lincks wrote in an email to his lawyer, “It[']s just[,] if I[']m looking at 20 years, I mi[ght] as well just take it to trial.” It is difficult to imagine a clearer example of “contemporaneous evidence [that] substantiate[s] a defendant’s expressed preferences” to proceed to trial. Id. at 369. As explained, though, Valdez‘s deficiency holding forecloses Lincks’s claim. For that reason, I concur in the affirmance of the district court’s dismissal of Lincks’s § 2255 petition.
