Finding Writer

Convert raw pentest notes, log snippets, vulnerability descriptions, or any unstructured observation into a complete, report-ready audit finding.

When NOT to use

Do NOT apply the finding format — respond normally — when:

• The user is asking a general security question, not describing a specific vulnerability they found
• The user wants to know how to exploit something (use the relevant hunter skill instead)
• The user is asking for a CVSS score only (use cvss-scorer instead)
• The user is asking how to fix a vulnerability (use remediation-planner instead)
• No evidence, observation, or artifact has been provided — ask for it rather than writing a finding from theory

Process

1. Parse Input: Identify the vulnerability class, affected component, and any evidence provided
2. Clarify if Needed: If the input is ambiguous or missing critical context, ask ONE clarifying question before proceeding — do not stall
3. Assess Severity: Score based on actual exploitability and impact in context, not just the vulnerability class — see references/severity-guide.md. Run scripts/cvss-scorer.py to produce the CVSS v3.1 vector and base score.
4. Tag the Finding: Run scripts/cwe-search.py <keyword> to look up the correct CWE ID and OWASP category for the vulnerability type.
5. Write the Finding: Fill every field using only information provided or derivable from the input
6. Flag Gaps: Mark any field without sufficient input as [TO BE ADDED]

Output Format

Title: Short, action-oriented title describing the vulnerability (e.g. "Unauthenticated Access to Admin Endpoint")

Severity: Critical / High / Medium / Low / Informational Justification: one sentence explaining the rating based on exploitability and impact

CVSS: CVSS:3.1/AV:_/AC:_/PR:_/UI:_/S:_/C:_/I:_/A:_ — [score] ([rating])

Description: What the vulnerability is, where it was found, and why it matters. 2–4 sentences, technical but readable by a non-specialist.

Evidence: The specific proof — URL, HTTP request/response snippet, log line, screenshot reference, or step-by-step reproduction steps. Be precise. If not provided: [TO BE ADDED]

Impact: What an attacker can achieve by exploiting this. Be concrete — data exposure, privilege escalation, account takeover, RCE, lateral movement, etc.

Recommendation: What to fix and exactly how. Not "fix the access control" — but "validate object ownership server-side on every request to /api/resource/{id} before returning data."

References: CWE ID, OWASP Top 10 category, and/or CVE if applicable. See references/owasp-cwe-map.md for common mappings.

Rules

• Never invent evidence — if the user didn't provide it, mark it [TO BE ADDED]
• Severity reflects actual exploitability and business impact in context, not the vulnerability class alone
• Recommendations must be specific enough that a developer can implement the fix without follow-up questions
• Ask at most one clarifying question per finding
• Do not add disclaimers, caveats, or legal boilerplate to the output
• See references/examples.md for a complete example at each severity level