security-talent/bugskill-ai

observe-skill

Logs the outcome of a skill execution to observations/ /runs.md.

소스 보기

사용 방법

  1. 설치 명령을 복사하고 자신의 에이전트 환경에 설치하기 전에 소스 패키지를 확인하세요.
  2. 필요한 도구, 인증 정보, 지원 환경은 원본 문서에서 확인하세요.
  3. 입력과 원하는 산출물을 설명하고 출처와 결과를 대조하세요.
npx skills add https://github.com/security-talent/bugskill-ai --skill observe-skill

시작 전 확인

  • 설치에는 모델 구독, API 한도, 파일 접근 권한이 포함되지 않습니다.
  • 별점 수와 설치 수는 활동 지표이며 독립적인 품질·보안 검증이 아닙니다.
원본 Skill 문서

원본 저장소의 제목, 예시, 코드, 표, 링크, 이미지를 유지해 표시합니다.

Observe Skill

Purpose

Track how skills perform in practice. A skill that works in theory may miss steps, use outdated payloads, or fire on the wrong triggers in real-world conditions. This skill captures that gap by recording execution outcomes to a per-skill log file.

Trigger Conditions

Activate this skill when the user says any of:

  • "log this run"
  • "skill worked" / "that worked"
  • "skill failed" / "that didn't work" / "this didn't work"
  • "log the outcome"
  • "record this run"
  • After any skill completes and the user describes a result (success, partial, fail)

Methodology

  1. Identify which skill was just used — ask if unclear. The skill name must match a directory name under skills/.
  2. Determine the outcome — success (goal achieved), partial (some steps worked, some didn't), or fail (skill produced no useful result).
  3. Collect details:
  • Date (today's date)
  • Task description (what was the user trying to do?)
  • What specifically worked (payloads, steps, tools that produced results)
  • What specifically failed (steps that produced no result, wrong output, or errors)
  • Any error messages or unexpected responses verbatim
  1. Write the entry — append to observations/<skill-name>/runs.md. Create the file and directory if they don't exist.
  2. Confirm — show the user the appended entry.

Entry Format

---
date: YYYY-MM-DD
skill: <skill-name>
task: <one-line description of what was attempted>
outcome: success | partial | fail
what_worked: |
  <describe what steps, payloads, or approaches produced results>
what_failed: |
  <describe what steps or payloads did not work>
errors: |
  <paste any error messages or unexpected responses>
notes: |
  <any additional context — target type, environment quirks, etc.>
---

Example Interaction

User: "log this run — bola-idor worked, I found a sequential integer ID on /api/orders and was able to read other users' orders using my own session token"

Response: Creates or appends to observations/bola-idor/runs.md:

---
date: 2026-03-14
skill: bola-idor
task: Test /api/orders endpoint for horizontal IDOR using sequential integer ID
outcome: success
what_worked: |
  Sequential integer IDs confirmed. Substituting victim's order ID (attacker ID +/- 1-10)
  returned full order details including name, address, and payment last4. Tested with
  attacker's Bearer token — no ownership check.
what_failed: |
  UUID endpoint /api/orders/uuid was not vulnerable — UUIDs not guessable.
errors: |
  None.
notes: |
  REST API, no GraphQL. /api/v1/ also vulnerable. /api/v2/ returns 403 on cross-user ID.
---

Fix Patterns

This is a meta-skill — it has no fix patterns. It generates data that other skills (especially /amend-skill) consume.

같은 저장소의 Skills

더 많은 Skills

모든 Skills