| name | program-intel |
| description | Summarize the historical bug patterns for a specific bug bounty program/team using the local disclosed-report corpus. Use when the user names a program (e.g. shopify, gitlab, nextcloud, nodejs, uber, tiktok) and wants to know what's been found there, recurring weak spots, hot endpoints/assets, and what to avoid duplicating. Helps focus hunting and avoid dupes. |
program-intel
Profile a program from its disclosed reports so the user knows where to look and
what's already been reported.
Workflow
-
Resolve the handle. Programs are identified by Handle in each report. List
what's available / find the closest handle:
grep -h "^- \*\*Handle\*\*:" reports/*.md | sort | uniq -c | sort -rn | grep -i "<name>"
High-coverage handles include: deptofdefense, ibb, nextcloud, security, shopify,
nodejs-ecosystem, gitlab, curl, x, vkcom, uber, shopify-scripts, tiktok, brave.
-
Pull the program's reports:
python3 .claude/skills/_lib/search_reports.py "<broad term or *>" --handle <handle> --limit 60
(Or to enumerate all of a program's reports, grep filenames after finding their
IDs.) Read the highest-severity / most-disclosed ones in full.
-
Synthesize an intel brief:
- Volume & severity mix — how many reports, breakdown by severity.
- Recurring vuln classes — what keeps getting found (rank by frequency).
- Hot assets / endpoints / features — subdomains, APIs, app areas that
repeatedly show up.
- Notable chains — high-impact reports worth studying, with
#<id>.
- Where to hunt next — under-tested adjacent surface, variations of past
bugs, classes that appear elsewhere but not yet here.
- Dupe risk — patterns already well-covered (likely to be dupes now).
Rules
- Cite report IDs and
reports/<file> for every claim.
- Distinguish "frequently found here" (mature, dupe-prone) from "found on similar
programs but not here" (opportunity).
- Note the date range of the reports — old patterns may be patched.