Bugcrowd Reporting

# BUGCROWD REPORTING — Program-Specific Tactics

> Companion to the generic report-writing skill. Use when working specifically on Bugcrowd submissions where VRT mapping, OOS-clause rebuttals, or per-program target selection matter.

This skill encodes patterns that apply specifically to Bugcrowd's submission flow. For the generic per-platform templates (HackerOne / Bugcrowd / Intigriti / Immunefi report bodies), use the report-writing skill. For the 7-Question Gate before deciding to report at all, use triage-validation.

Bugcrowd's submission form requires a single VRT (Vulnerability Rating Taxonomy) selection. The dropdown's default severity is bound to the chosen node — pick wrong and the form auto-suggests P4 when the actual impact is P3 or P2.

1.1 Search hierarchy (try in order, pick the highest-severity match that still describes the bug)

For any finding, search the VRT dropdown with these terms in this order:

• The bug's primary class — e.g., IDOR, XSS, SSRF, auth bypass, 2FA bypass
• The data category exposed — e.g., PII, sensitive data exposure, disclosure of secrets
• The control bypassed — e.g., broken access control, authentication bypass
• The endpoint type — e.g., no rate limiting on form > login, no rate limiting on form > change password
• The generic parent node — e.g., Server Security Misconfiguration > Other, Broken Access Control > Other

1.2 Pick the highest-severity match that still accurately describes the bug

Never select a VRT that misrepresents the bug just to get a higher default severity. Triagers will reassign and may flag the misrepresentation. The discipline is: pick the most specific accurate VRT, then use §2 (Manual Severity Override) if the default is wrong.

1.3 Common mappings worth knowing

1.4 If no good VRT exists

Pick Server Security Misconfiguration → Other or Broken Access Control → Other and lead the description body with a "VRT mapping note" explaining why the chosen node is the closest available match and what the bug actually is.

Bugcrowd's form lets you manually set Technical Severity separate from the VRT default. The form text itself states: "A severity rating suggested by the VRT is not guaranteed to be the severity rating applied to your submission."

2.1 When to override

Override the VRT default when:

• The chained outcome is a higher severity than the standalone bug class (chain → ATO)
• The VRT category is approximate and its default doesn't reflect the actual impact class
• The program's own Focus Areas explicitly list this outcome at a higher severity than VRT's default
• The data class exposed is more sensitive than the VRT's example uses (e.g., real-name PII vs. handle enumeration)

2.2 How to override (form mechanics)

• Select the VRT that most accurately describes the bug (per §1)
• Note the auto-suggested severity (P4, P3, etc.)
• In the Technical Severity field, manually pick the severity you're requesting
• Add a Severity Request paragraph as the FIRST section of the description body (per §3)

2.3 Don't over-claim

P1 inflation is the fastest way to lose triager trust. Reserve P1 for ATO without interaction, RCE, mass PII exfiltration, fund theft, and similar Critical-bucket impacts. If the chain to P1 requires a separate stolen-cookie premise, file the standalone primitive at P3 and discuss the chain explicitly with cross-references (per §4).

When the VRT default underrates impact, the FIRST section of the description body should be a severity-request paragraph. This is the first thing the triager reads and it pre-empts the auto-close that often happens when triagers see "P4" in the form.

3.1 Template

```markdown

The closest VRT category for this finding is "[chosen VRT]," which Bugcrowd defaults to P[N] ($X-$Y in [program]'s rubric). I am requesting evaluation at P[M] [standalone | in chain with submission XXXX] for the following reasons:

• [Impact axis 1] — [specific reason this finding exceeds the VRT default's example]
• [Impact axis 2] — [specific reason, often citing the program's own Focus Areas]
• [Impact axis 3] — [specific reason, often comparing to historical bounty payouts for the same data class]

### 3.2 Worked example pattern

The closest VRT category for this finding is "No Rate Limiting on Form → Login," which Bugcrowd defaults to P4 ($XX–$YY in this program's rubric). I am requesting evaluation at P3 standalone, P2 in chain with submission XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX for the following reasons:

• [Endpoint] is a password-disclosure oracle, not a generic rate-limit gap. The endpoint returns true/false for a supplied password — combined with cookie theft, it is the exact primitive an attacker uses to learn the victim's durable credential from a stolen transient session.
• The chained outcome is "Account Takeover," explicitly listed under the program → Focus Areas → Critical (P1) examples. This finding is one of two independent links that make that chain work; the other is submission XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX.
• The two findings have independent fix surfaces. Per the program's "one fix = one bounty" rule they should be evaluated as two separate findings whose chain raises the joint severity.

3.3 Tone guidelines

• Open with "Severity request — please review carefully before applying VRT default" (literal phrase). Triagers recognize this and don't auto-close.
• Use bold sparingly; one or two phrases per reason at most.
• Never use "could potentially" or "may allow" — these are downgrade-bait.
• Cite the program's own Focus Areas / accepted-impact list by exact name.
• Cross-reference linked submissions by full submission ID (UUID format).

Bugcrowd triagers sometimes auto-close findings by mapping them to an OOS clause without reading the report. Pre-empt this by including an In-scope justification section that quotes the OOS clause and explains why your finding doesn't fit it.

4.1 "Rate limiting on non-authentication endpoints" rebuttal

When the finding is on an authentication endpoint (login, password change, password verify, OTP verify, token validate):

```markdown

The program OOS list excludes "Rate limiting or brute-force issues on non-authentication endpoints." [endpoint name] is the canonical authentication endpoint — it accepts a password/token/OTP and returns whether it matches the user's stored credential. By definition this is an authentication primitive. The OOS clause's "non-authentication" qualifier therefore does NOT apply.

 ### 4.2 "Debug information disclosure" rebuttal (for schema/introspection findings)

When the finding involves schema disclosure that's actually a control bypass (e.g., GraphQL allowlist bypass exposing the mutation surface):

This is not a "debug information" finding. The [persisted-query allowlist | introspection gate | etc.] is acting as an authorization control on the [GraphQL surface | API tree | etc.]. The bypass converts that gate into a no-op — multiple [mutations | endpoints] become reachable that the official clients cannot invoke. The schema disclosure is incidental; the impact is mutation-surface unlock, not information leakage.

 ### 4.3 "User enumeration with low-risk information" rebuttal

When the finding leaks more than just "this handle/email exists":

The program OOS list excludes "User Enumeration issues with low-risk and insignificant information being enumerated." The data leaked here is neither low-risk nor insignificant: it includes the matched user's [full real first name, last name, profile photo / SSN last 4 / phone number / etc.]. Real-[name | identity | etc.] knowledge is the same data class that gates the program's own [account-recovery flows / fraud-investigation interviews / support-call verification], which means this lookup directly defeats one of the program's own anti-fraud controls. This is PII disclosure, not handle enumeration.

 ### 4.4 "Theoretical issues / not exploitable" rebuttal

When the finding's exploitability has been demonstrated end-to-end:

This finding is exploitable end-to-end as demonstrated in the PoC: [one-sentence summary of the proven exploitation path]. The PoC includes [N HTTP requests with redacted cookies | N screenshots | a sanitized HAR file | etc.] showing the attacker's session [reading victim data | changing victim credentials | etc.]. The OOS clause for "theoretical issues that are not exploitable, or can not be demonstrated as exploitable" does not apply. ```

4.5 If you don't have end-to-end proof

If you only have an "API behavior" observation without a fully demonstrated exploitation path, the finding is theoretical. Don't file it. Capture the missing evidence (delivery confirmation, token-format inspection, cross-account victim data in response, etc.) and re-evaluate. Filing theoretical findings damages researcher reputation on the platform.

Many high-impact findings are chains of two or more standalone primitives (oracle + missing step-up = ATO). Bugcrowd's program rule: "Submit one vulnerability per report. If multiple vulnerabilities must be chained to demonstrate impact, you may include them together, but please clearly explain the chain."

5.1 Filing strategy for chains

• Identify the highest-severity chained outcome (typically the one that maps to a P1/P2 example in the program's Focus Areas).
• File the chain consumer FIRST — the report whose body describes the full ATO/RCE/etc. impact. Use the chained severity (e.g., P1).
• Within minutes, file the chain primitives as separate reports — each at its standalone severity (typically P3/P4) but cross-referencing the consumer.
• In each primitive's body, explicitly note: "Chain partner: submission [UUID] for the full impact narrative."

5.2 Cross-reference template

In the chain consumer's body, near the top:

```markdown

• submission [UUID-1] — [primitive 1: e.g., me.verify_password no rate limit]
• submission [UUID-2] — [primitive 2: e.g., GraphQL APQ allowlist bypass]

In each primitive's body:

This finding is the [N]th of [M] independent primitives that together produce [impact, e.g., "permanent account takeover from a stolen session cookie"]. The full chain narrative and the chained severity (P1) are documented in submission [UUID]. This standalone report is at P3 reflecting the primitive's individual impact. ```

5.3 What NOT to do

• Don't paste the entire chain narrative into every primitive's body. Cross-reference instead.
• Don't claim each primitive is independently P1 — overclaiming damages credibility for all linked reports.
• Don't ask for a single combined bounty for the chain. The program's "one fix = one bounty" rule means each independent fix surface is a separate bounty. Frame the chain as a severity-amplifier, not a merge request.

Programs that scope both production and QA assets (e.g., app.target.com vs. app-qa.target.com) require careful target selection.

6.1 Picking the right target dropdown entry

6.2 QA-disclaimer paragraph (when filing a QA-tested finding)

```markdown

• Tested in the QA environment ([qa-host]) per the program's instruction for non-US researchers / per the program's QA-testing scope.
• Production may enforce additional controls that QA does not — please confirm whether the same resolver code path runs on production.
• All testing limited to my own QA test account; no other accounts were interacted with.
• Selected target: [chosen-target]. Selected VRT: [chosen-VRT]. Both selected as closest available matches; please reassign if more appropriate options exist.

These are program-rule items that affect submission quality even though they're not part of the report body itself.

7.1 Account email — Bugcrowdninja alias

Most Bugcrowd-managed programs ask researchers to use their @bugcrowdninja.com alias as the account email on the target. This flags traffic as friendly-tester to the target's fraud team. Steps:

• Sign up for the target using your Bugcrowd alias (looks like username@bugcrowdninja.com)
• If you already have an account, update the email to the alias before testing
• If the program offers a contact path for multi-account testing, use that to coordinate additional aliased accounts (alias+test1@bugcrowdninja.com, etc.)
• Document the alias in the report's "Notes for triagers" section so the triager can correlate test traffic in their logs

7.2 Account-state restoration after PoC

After demonstrating any state-changing action (password change, email change, device removal, money transfer):

• Restore the account to its pre-PoC state immediately (e.g., revert the password to what it was)
• Document the restoration in the report — "Account state was restored immediately after the PoC."
• If restoration is impossible (e.g., a sent email cannot be unsent), document the residual state honestly

7.3 Cookie / session rotation after submission

After submitting any report that includes session-cookie-derived evidence:

• Log out and log back in to rotate the session cookie
• Any cookie value visible in screenshots / HARs you submitted is now dead
• Optionally, rotate the test-account password to a fresh value not used elsewhere — that way passwords visible in PoC screenshots are also dead

7.4 If the test account gets locked

• Don't create a second account to bypass the lock — that looks like fraud
• Email the program-specific contact (if provided) with your Bugcrowd username and the report ID
• Wait for the program team to unblock you; testing during the lock period damages credibility

When you have multiple findings from a single engagement, the order matters.

8.1 Recommended order

• The highest-severity, best-evidenced finding first. Locks in your timestamp on the most valuable bug. If it's a chain consumer, follow with the chain primitives within an hour.
• Cross-referenced primitives next. File each within an hour of the chain consumer so the cross-references are tight.
• Standalone P3 findings next. Order by quality of evidence: best-evidenced first, riskier (OOS-adjacent) findings later.
• OOS-risky findings last. By the time these land, you've established a credible track record on this engagement.

8.2 Tracking submission IDs

Maintain a simple text file with each submission's UUID, severity, and one-line description. You'll reference these IDs across reports and during any subsequent triager dialogue.

8.3 What to avoid

• Don't file all findings in a single batch within minutes. Triagers see this as low-effort spam.
• Don't file an OOS-risky finding before a clean P3 from the same engagement. The clean one establishes credibility; the risky one benefits from being read in that context.
• Don't disclose any findings publicly until the program explicitly says it's OK. Bugcrowd's confidentiality applies to both unresolved AND resolved issues.

This skill is small and focused. It does NOT duplicate content from report-writing (per-platform templates, CVSS scoring, downgrade counters, 60-second pre-submit checklist) — load both skills together when filing a Bugcrowd report. The two skills' content composes naturally because their boundaries are clean: report-writing is the body templates, bugcrowd-reporting is the program-specific tactics layered on top.

• report-writing — Always load this skill alongside bugcrowd-reporting. Workflow primitive: report-writing provides the body skeleton; this skill layers VRT selection + severity-request paragraph + OOS-clause rebuttals on top of that skeleton.
• triage-validation — When deciding if a Bugcrowd-bound finding will pass triage. Workflow primitive: 7Q gate runs first; only validated findings reach this skill for VRT mapping.
• evidence-hygiene — When attaching screenshots / HARs to a Bugcrowd submission. Workflow primitive: Bugcrowd's private attachment system requires redacted evidence; route everything through evidence-hygiene before clicking attach.
• security-arsenal — When the PoC step needs payloads cited verbatim. Workflow primitive: the "Steps to Reproduce" section in this skill's report bodies pulls exact payloads from security-arsenal so the triager can paste-and-run.
• bb-methodology — When confirming engagement mode is bug-bounty (not red-team). Workflow primitive: PART 0 of bb-methodology answers "bug bounty?"; if yes AND the target is Bugcrowd, this skill becomes the reporting overlay.