h1-reports
Query a local corpus of 14,878 disclosed HackerOne reports (program, title, link, bounty, upvotes, vuln_type) from reddelexc/hackerone-reports. Use for PROGRAM-SPECIFIC bounty calibration ("what does
SKILL.mdskills/h1-reports/SKILL.md
GitHubh1-reports — Program-Specific HackerOne Payout Corpus
Companion to bb-disclosures (multi-platform severity corpus) and h1-playbooks (per-class attack-pattern playbooks distilled from this same corpus — read those when planning a hunt for a specific class). This one's unique strength: program names + upvotes, so it answers "what does THIS program historically pay for THIS bug class?"
Queries
python <skill_dir>/scripts/query.py --program gitlab --vuln idor --min-bounty 1000
python <skill_dir>/scripts/query.py --vuln "business logic" --min-bounty 5000 --limit 20
python <skill_dir>/scripts/query.py --q "graphql" --program github
python <skill_dir>/scripts/query.py --stats # top programs/vuln types by total payout
python <skill_dir>/scripts/query.py --refresh # pull latest data.csv from GitHub
Columns: program, title, link, upvotes, bounty, vuln_type (links are hackerone.com/reports/N — read the full writeup for escalation patterns when a record is relevant).
Use it in the hunt workflow
- Before reporting any finding (with bb-disclosures): query
--program <target> --vuln <class>→ if the program appears, its own history is the strongest severity argument available. If not, query--vuln <class> --min-bounty <expected>for cross-program precedent. - When scoping a new program:
--program <name> --sort bountyshows what that program pays for — tells you which hunting skills to lead with (IDOR-heavy → bb-authz/idor-runner; business-logic-heavy → bb-bizlogic). - For escalation ideas on a confirmed primitive:
--vuln <class> --min-bounty 5000→ read the top-paid writeups for that class; chains that reached 4-5 figures show the missing links worth adding to your own finding.
Discipline
- Payout history is an argument, not a guarantee — pair with CVSS and program policy, never substitute.
- Bounty
0.0usually means disclosed-but-unrewarded (or reward hidden); don't treat as "program pays nothing". - Corpus snapshot ages;
--refreshbefore any serious calibration session.