bugcrowd-reporting skill

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.

1. VRT Category Selection — Search & Fallback Strategy

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 a lower priority (often P4) when the actual impact is P3 or P2.

Note: VRT default severities are not fixed constants. Bugcrowd revises the VRT schema across versions, and individual programs can remap defaults via their own priority configuration. The P-values shown in the examples below (e.g., "No Rate Limiting on Form → Login" defaulting to P4) are the typical baseline at time of writing — always read the severity the current form actually auto-suggests for this program rather than assuming the value here.

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:

1. The bug's primary class — e.g., IDOR, XSS, SSRF, auth bypass, 2FA bypass
2. The data category exposed — e.g., PII, sensitive data exposure, disclosure of secrets
3. The control bypassed — e.g., broken access control, authentication bypass
4. The endpoint type — e.g., no rate limiting on form > login, no rate limiting on form > change password
5. 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.

2. Manual Severity Override

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)

1. Select the VRT that most accurately describes the bug (per §1)
2. Note the auto-suggested severity (P4, P3, etc.)
3. In the Technical Severity field, manually pick the severity you're requesting
4. 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).

3. Severity-Request Paragraph — Always First in the Body

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

## Severity request — please review carefully before applying VRT default

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:

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

[Optional: one paragraph contrasting the VRT default's typical example with this finding's actual impact, e.g., "P4 = No Rate Limiting on Login Form" applies to ordinary login pages where brute force is bounded by lockout policies; this endpoint has neither lockout nor throttle.]

3.2 Worked example pattern

## Severity request — please review carefully before applying VRT default

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:

1. **[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.
2. **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.
3. **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.

The "P4 = No Rate Limiting on Login Form" default applies to ordinary login pages where brute

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).

4. OOS-Clause Rebuttals — Templates

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):

## In-scope justification

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):

## In-scope justification

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":

## In-scope justification

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:

## In-scope justification

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.

5. Chained Findings — Cross-Reference Strategy

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

1. Identify the highest-severity chained outcome (typically the one that maps to a P1/P2 example in the program's Focus Areas).
2. File the chain primitives FIRST as separate reports — each at its standalone severity (typically P3/P4). A submission's UUID only exists after it is filed, so the primitives must exist before anything can reference them. Leave the cross-reference line in each primitive's body as a placeholder for now.
3. File the chain consumer — the report whose body describes the full ATO/RCE/etc. impact at the chained severity (e.g., P1). Because the primitives already exist, fill in their real UUIDs in the consumer's "Chain partners" block (per §5.2).
4. Edit each primitive after filing the consumer to backfill the consumer's UUID: "Chain partner: submission [UUID] for the full impact narrative." (The reference is bidirectional, so one report must be edited after the other is filed — Bugcrowd allows post-submission edits to the description body.)

5.2 Cross-reference template

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

## Chain partners (filed as separate reports)

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

These primitives have independent fix surfaces and are filed separately per the program's "one fix = one bounty" rule. The chained impact described here requires both primitives to be exploited end-to-end.

In each primitive's body:

## Chain context

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.

6. Target Selection — QA vs. Production

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