Pwnote Bug Bounty

Bug Bounty Workflow

Reference for running a bug bounty engagement end-to-end: program intake, recon, vulnerability classes, report writing, and triage disputes. Output should be structured so it can be pasted directly into a pwnote engagement file (see the pwnote-engagement-file skill if present) — findings map to the findings array, recon steps map to docs/blocks.

1. Program Intake

When given a program's scope page (or a pasted policy), extract into a structured note:

| Field | Notes | |---|---| | In-scope assets | domains, IP ranges, app names, mobile app IDs, source repos | | Out-of-scope | explicit exclusions — always check for staging/subdomains carved out | | Reward table | severity → payout, so you can prioritize effort against expected value | | Special rules | rate limits, no-automated-scanning clauses, social engineering restrictions, disclosure timeline requirements | | Safe harbor language | confirm it exists before doing anything invasive |

Always re-read scope before starting active testing — programs change scope frequently and testing out-of-scope assets voids safe harbor.

2. Recon Workflow

Standard funnel, in order:

1. Asset discovery — subdomain enumeration (passive + active), ASN/IP range expansion, cloud bucket discovery
2. Live host probing — resolve + HTTP probe, screenshot triage for anomalies
3. Tech fingerprinting — stack identification to prioritize known-CVE-prone components
4. Content/endpoint discovery — JS file parsing for endpoints/secrets, historical URL sources (archive snapshots), directory brute-forcing within scope/rate limits
5. Asset triage — rank by attack surface (auth surfaces, file upload, admin panels, API endpoints with IDs in the path) rather than testing everything uniformly

Log each recon step as a command-output or host block per target so the trail is reproducible later for the report.

3. Vulnerability Class Playbooks

Keep class-specific methodology in references/ rather than inline here — load only the relevant one:

• references/vuln-classes.md — IDOR/BOLA, SSRF, auth/session bypass, business logic abuse, race conditions: what to check for and how to structure the finding, at a methodology level (not payload cookbooks)
• references/report-templates.md — per-platform report formatting differences (H1 vs Bugcrowd vs Intigriti vs YesWeHack)

4. Report Writing

Every report needs, regardless of platform:

• Title — specific and severity-signaling ("IDOR on /api/v1/orders/{id} allows viewing other users' orders", not "IDOR vulnerability")
• Summary — one paragraph, plain language, what an attacker can do
• Steps to reproduce — numbered, minimal, reproducible by someone with zero context
• Impact — concrete, tied to the program's asset (data exposure scope, account takeover chain, financial impact)
• PoC — request/response pairs or a short script; redact any real user data
• Remediation — specific enough to be actionable, not "add input validation"
• CVSS — vector + score if the program asks for it; be conservative, don't inflate

Use references/report-templates.md for the platform-specific field layout.

5. Triage & Dispute Handling

If a program pushes back (duplicate claim, N/A, severity downgrade):

• Ask for the duplicate report reference/timestamp before conceding — programs sometimes cite dupes that don't actually match your finding's root cause
• If severity is downgraded, restate impact in terms of the program's own asset criticality rather than re-arguing CVSS abstractly
• Keep disputes factual and low-heat; escalation to platform mediation is a last resort, not a first move

6. Output Mapping (pwnote)

When generating structured output for a pwnote import:

• One finding object per confirmed vuln (title, severity, cvssScore, cvssVector, cwe, affectedAssets, description, poc, remediation, references, status)
• Recon steps as host/command-output blocks in a 01_Recon doc
• Exploitation chain as code/http-request blocks in a 02_Exploitation doc
• Use a topology or killchain board if the finding involves chaining multiple assets/creds

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Pwnote
• Source: Pwnote/skills
• License: MIT

Install and usage instructions live in the source repository linked above.