MESKILL, Circuit Judge:

This is an appeal from a judgment of conviction entered in the United States District Court for the Southern District of New York, Kram, J., following a jury trial. Defendant-appellant Louis Cosentino was convicted of extortion in violation of 18 U.S.C. § 1951 (1982) and use of the mails to facilitate bribery in violation of 18 U.S.C. §§ 2, 1341 (1982). Two other counts were dropped prior to trial. Cosentino challenges the admission of witness cooperation agreements during direct testimony of government witnesses and claims that certain questions and comments by the prosecutor constituted prejudicial misconduct. For reasons that follow, we affirm.

BACKGROUND

Defendant-appellant Cosentino was a Project Superintendent for the New York City Housing Authority. As a superintendent, he had limited authority to purchase materials not available through the Housing Authority by placing "certificate for payment" orders with private vendors. Although Authority rules prohibited superintendents from placing more than $500 in such orders with any single vendor, Cosentino allegedly evaded this limitation by splitting orders among multiple companies owned by each vendor. He allegedly employed the same stratagem to circumvent another Authority rule that proscribed placement of more than one order with any single vendor in any thirty day period. According to the government, Cosentino solicited and received kickbacks from the vendors with whom he placed certificate for payment orders.

At trial, the government's case rested almost exclusively on the testimony of Alan Rappaport and Irving Eisenberg, two vendors who had dealt with Cosentino. They testified about the kickback scheme, explaining that Authority superintendents in general and Cosentino in particular placed orders only with vendors who kicked back a percentage of each order to the superintendent. There was also testimony that Cosentino received a $1,000 loan that he "worked off" by placing $10,000 worth of certificate for payment orders without demanding his usual ten percent kickback.

Cosentino took the stand on his own behalf to explain that he had split orders only in order to obtain necessary supplies that could not readily be procured through the Authority. He had evaded the rules, he said, to provide better service to the projects under his supervision. He also explained that he had paid the loan back out of his own pocket and emphatically denied soliciting or accepting bribes.

Because the case against Cosentino depended so heavily on the testimony of Rappaport and Eisenberg, their credibility was the central battleground of the trial. As participants in the kickback scheme who had agreed to testify in return for guilty pleas on reduced charges, they were especially vulnerable to impeachment. As a result, both the prosecuting Assistant United States Attorney (AUSA) and Cosentino's counsel highlighted credibility issues in their respective opening statements to the jury.

In opening, the prosecutor outlined the case against Cosentino, alluding specifically to the witnesses' background and cooperation agreements as follows:

I will tell you now about Rappaport and Eisenberg, the two vendors you will hear testify in this trial. They were not innocent victims. They acknowledge their participation in corrupt and criminal [a]ctivity. They pleaded guilty to felony charges and have been sentenced, but before they pled guilty they entered into a cooperation agreement with the U.S. Attorney's Office, an agreement to [sic] which the U.S. Attorney's office agreed to accept their pleas and in return they agreed to testify.

Remember, people who engage in conspiracy don't act in the open, and, as the evidence in this case will show, they tried not to keep records of their dealings. They acted in secret.

So for us to be able to show you what happened at Highbridge Houses [where Cosentino was superintendent], it's necessary that we bring people right from out of the muck to testify before you today.

Because Rappaport and Eisenberg participated in corrupt activities and because they have an agreement with the government, the government asks you to scrutinize their testimony very carefully. You may not like a lot of the things they have done, but, remember, they are not on trial here today. Only Louis Cosentino is on trial, so when Alan Rappaport and Irving Eisenberg testify, you should be asking yourself one question: Are they telling you the truth today or tomorrow?

Listen closely to what they say and see if what they say doesn't make sense in light of all the evidence in this case.

Tr. 11-12. Cosentino's counsel made a brief opening statement devoted almost entirely to the credibility issues. He focused on specific aspects of the cooperation deals that Rappaport and Eisenberg had struck with the government.

The only evidence they [ i.e., the government] are going to give you are the words of Alan Rappaport and Irving Eisenberg, and, as [prosecuting AUSA] Mr. Richman says to you, they are raised up from the muck.

The government says criminals will be testifying against the innocent man sitting at the defense table, Mr. Louis Cosentino.

You will find out that they are men who received large amounts of money from the New York City Housing Authority. You will find out that these are men who are convicted felons; both pled guilty in the court to felonies.

Cooperation agreements, they made deals with the government. A deal basically means: You give us what we want and we'll give you what you want.

What the government wants is a conviction.

So what are they going to do but come here and try to give the government what they want in exchange for which they will get lesser charges than originally exposed to, and, more importantly, their back is scratched by the government in a letter sent to the judge later on.

Mr. Eisenberg has already been sentenced, but if it were found out he did not give the government what the government wanted here in contradiction to what he said earlier, he leaves himself open to all the rest of the charges that he could have been charged with, and the government will go after him.

Tr. 13-14. During the direct testimony of both Rappaport and Eisenberg, the government offered the full text of their written cooperation agreements. The district court admitted them over defense objection. The jury requested both agreements, in addition to other exhibits and testimony, during its deliberations leading to Cosentino's conviction on both counts.

On appeal, Cosentino principally argues that the admission of the cooperation agreements during direct examination constituted impermissible bolstering of the witnesses' credibility. He also argues that the agreements were misleadingly incomplete and that a limiting instruction should have been given. Finally, he contends that certain questions and statements by the AUSA constituted prosecutorial misconduct. We reject these contentions and affirm Cosentino's convictions.

DISCUSSION I.

This appeal requires us to define the circumstances in which witness cooperation agreements may properly be admitted into evidence during the direct testimony of government witnesses. The existence and contents of such agreements are inevitably of considerable interest to both prosecution and defense. They tend to support witnesses' credibility by setting out promises to testify truthfully as well as penalties for failure to do so, such as prosecution for perjury and reinstatement of any charges dropped pursuant to the deal. The agreements can impeach, however, by revealing the witnesses' criminal background. Defense counsel can also argue that such witnesses cannot be believed because they are under pressure to deliver convictions and correspondingly tempted to twist facts to do so.

Cooperation agreements accordingly demand careful treatment under principles governing attack on and rehabilitation of witnesses' credibility. It is well settled that absent an attack, no evidence may be admitted to support a witness' credibility. See generally McCormick on Evidence § 49 (E. Clearly 3d ed. 1984); 3 J. Weinstein M. Berger, Weinstein's Evidence ¶ 608[08] (1987). We have invoked this rule in considering the admissibility of cooperation agreements because of their tendency to support or bolster credibility. See, e.g., United States v. Smith, 778 F.2d 925, 928 (2d Cir. 1985); United States v. Borello, 766 F.2d 46, 56 (2d Cir. 1985); United States v. Edwards, 631 F.2d 1049, 1051 (2d Cir. 1980); United States v. Arroyo-Angulo, 580 F.2d 1137, 1146 (2d Cir), cert. denied, 439 U.S. 913, 99 S.Ct. 285, 58 L.Ed.2d 260 (1978).

A witness' credibility is often tested by a sequence of attack on cross-examination followed by rehabilitation on redirect. It may sometimes be useful, however, to develop impeaching matter in direct examination of a "friendly" witness in order to deprive an adversary of the psychological advantage of revealing it to the jury for the first time during cross-examination. We have accordingly held that impeaching aspects of cooperation agreements may be brought out in the government's direct examination of a witness who testifies pursuant to such an agreement. See Borello, 766 F.2d at 57; Edwards, 631 F.2d at 1051-52. Cf. United States v. Fernandez, 829 F.2d 363, 365 (2d Cir. 1987) (discussing scope of permissible reference to agreement in direct examination). Even in the absence of a prior attack on credibility, "the elicitation of the fact of the agreement and the witness' understanding of it, as a motivation for the witness to testify for the Government, should be permitted on direct examination in order to anticipate cross-examination by the defendant which might give the jury the unjustified impression that the Government was concealing this relevant fact." Edwards, 631 F.2d at 1052.

Because of the bolstering potential of cooperation agreements, however, we have permitted such agreements to be admitted in their entirety only after the credibility of the witness has been attacked. See Smith, 778 F.2d at 928. This restriction proceeds from our view that "the entire cooperation agreement bolsters more than it impeaches." Edwards, 631 F.2d at 1052; see also Borello, 766 F.2d at 56-57. Thus, although the prosecutor may inquire into impeaching aspects of cooperation agreements on direct, bolstering aspects such as promises to testify truthfully or penalties for failure to do so may only be developed to rehabilitate the witness after a defense attack on credibility.

Such an attack may come in a defendant's opening statement. If the opening sufficiently implicates the credibility of a government witness, we have held that testimonial evidence of bolstering aspects of a cooperation agreement may be introduced for rehabilitative purposes during direct examination. See Smith, 778 F.2d at 928; United States v. Jones, 763 F.2d 518, 522 (2d Cir.), cert. denied, 474 U.S. 981, 106 S.Ct. 386, 88 L.Ed.2d 339 (1985); United States v. Maniego, 710 F.2d 24, 27 (2d Cir. 1983) (per curiam). In such a situation the "rehabilitation" stage has already been reached on direct.

As a threshold matter, we must first decide whether the government's opening statement permissibly referred to the agreements in this case. Consistent with the foregoing principles, a prosecutor may refer to a witness cooperation agreement in opening only to the extent he or she could develop the same matter in direct questioning of a witness whose credibility has not been attacked. The prosecutor thus may advert to the existence of the agreement and related impeaching facts such as the witness' criminal background, but may not raise bolstering aspects of the agreement such as truth telling provisions, charge reinstatement conditions or penalties of perjury clauses. Because the witness alone can testify to his or her understanding of the agreement, the prosecutor also cannot discuss this aspect in opening. The excerpt from the record set forth above makes plain that the government in this case properly restricted its opening to appropriate matters.

It is also plain that Cosentino's counsel sufficiently raised matters of credibility in opening that the government could develop the whole cooperation agreements on direct. Cosentino's counsel made representations about reduction of charges and government intervention on behalf of the witnesses in the sentencing process, as well as the possibility of prosecution for charges dropped in exchange for testimony in the event the witnesses failed to deliver convictions. These references clearly opened the door to rehabilitation on direct by evidence of bolstering aspects of the cooperation agreements.

We thus arrive at the central issue in this appeal: if evidence of the whole agreement was admissible during direct examination, was it error to admit the agreement itself? We have not before squarely confronted a situation involving introduction of a whole cooperation agreement on direct following a sufficient attack on credibility in the defense's opening. Cosentino argues that our prior decisions establish a rule that cooperation agreements can be admitted in their entirety only on redirect, thus controlling the outcome of this case. We do not read those decisions so mechanically. It is true that some cases have addressed only the admission of testimony of bolstering aspects on direct examination. See Jones, 763 F.2d at 522; Maniego, 710 F.2d at 27. It is also true that in summarizing these cases and others, we stated in Smith that a prosecutor may "elicit testimony on the truth-telling portions . . . on direct," and may "introduce the entire agreement . . . on redirect examination to rehabilitate the witness." 778 F.2d at 928 (emphasis added). But the result in these cases is not, as Cosentino argues, the product of a distinction between direct and redirect examination. Rather, the event that renders the bolstering aspects of cooperation agreements admissible, on direct or redirect, is the attack on credibility that gives the bolstering evidence a rehabilitative purpose.

We hold that the written text of a cooperation agreement may be admitted during a witness' direct testimony whenever a defense attack on credibility in opening has made evidence of the whole agreement admissible. We see no reason to distinguish between the written text of the agreement and testimony about it if the rehabilitation stage has otherwise been reached by the time direct examination of the witness begins. The decision about the form evidence of the agreement should take lies within the trial judge's discretion under Fed.R.Evid. 403. In the exercise of that discretion, it may sometimes be appropriate to redact the agreement to eliminate potentially prejudicial, confusing or misleading matter. See, e.g., Arroyo-Angulo, 580 F.2d at 1145 n. 9 (deleting references to protective custody afforded witness' family); United States v. Koss, 506 F.2d 1103, 1112-13 n. 8 (2d Cir. 1974) (deleting references to organized crime and threats against witness' life on account of testifying), cert. denied, 421 U.S. 911, 95 S.Ct. 1565, 43 L.Ed.2d 776 (1975). No such redaction was necessary in this case. We conclude that there was no abuse of discretion in the district court's decision to admit the entire witness cooperation agreements during the direct testimony of the government witnesses.

II.

Cosentino's two remaining claims relating to the cooperation agreements merit only brief discussion. First, he argues that the agreements were misleadingly incomplete in that they omitted elements of the deals that would have undermined the witnesses' credibility. Specifically, Cosentino points out that each agreement stated that "[n]o additional promises, agreements and conditions have been entered into other than those set forth in this letter and none will be entered into unless in writing and signed by all parties," but neither agreement specified the charges dropped or mentioned that the witnesses were promised a choice among sentencing judges.

In general, there is little danger that significant terms will be omitted from cooperation agreements: the government and the defendant have considerable interest in seeing to it that the agreement spells out all terms of the bargain. Even though some minor details may occasionally be left out, the probative value of individual agreements, including their completeness, can most efficiently be evaluated by the trial judge in the exercise of discretion under Fed.R.Evid. 403. In the instant case, nothing suggests that that discretion was abused. Although the agreements apparently omitted some incidental matters, they contained the essential impeaching and bolstering information that Rappaport and Eisenberg had committed crimes, pled guilty and struck deals under which the government traded a degree of conditional prosecutorial lenience for "truthful" testimony. In any event, defense counsel brought out every one of the omitted matters on cross-examination and discussed them further in summation. Revelation of the other details in written form would have given Cosentino little added leverage in attempting to raise doubts in the jurors' minds about the witnesses' credibility. Under these circumstances, the decision to admit the full agreements had no adverse effect on Cosentino's defense and fell well within the ambit of sound discretion.

Second, Cosentino argues that a limiting instruction should have been given on the truth telling portions of the agreements. He seeks an instruction under which the jury could use the impeaching motive or bias aspects of the agreements to undercut credibility, but could not use the bolstering portions as support for credibility.

No request for such an instruction was made below, so we review only for plain error. See Fed.R.Crim.P. 30, 52(b). We discern no error of any kind in this respect. The proposed instruction would have been fundamentally at odds with the distinctions we have drawn between the impeachment and bolstering aspects of cooperation agreements. Although we have required a prior attack on credibility so that the whole agreement serves a rehabilitative function, we have never restricted use of an agreement to support credibility once that condition is satisfied. Our view presupposes that the agreement may and will be used to support credibility. The district court's treatment of the cooperation agreements was fully consistent with these principles.

III.

Cosentino raises three instances of allegedly improper questioning or comment by the prosecutor. He argues that the government improperly vouched for Rappaport's veracity by inquiring on redirect into his cooperation in other prosecutions. He also contends that the prosecutor improperly trapped him into stating that Rappaport and Eisenberg had lied in their testimony about the loan. He finally argues that certain rebuttal comments by the prosecutor in closing might have suggested to the jury that it could convict him without considering evidence of his good character on the issue of guilt.

We have considered these contentions and find them to be without merit. None warrants extended discussion.

CONCLUSION

For the foregoing reasons, Cosentino's conviction is affirmed.