authenticated-web-access
Reuse an existing browser login for bounded authenticated website requests when the user asks to read site data, download a file, inspect a network-backed app, or make a separately approved change. Only Firefox cookie acquisition is documented. Not for generic API authentication, password retrieval, or browser automation.
- License 许可
- MIT
- 所属仓库
- OrestesK/omp
- 最近更新
- 2026-10-16
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- authenticated-web-access
- description
- Reuse an existing browser login for bounded authenticated website requests when the user asks to read site data, download a file, inspect a network-backed app, or make a separately approved change. Only Firefox cookie acquisition is documented. Not for generic API authentication, password retrieval, or browser automation.
02 资源文件 1
- 参考文档references/firefox.md2.4 KB
03 SKILL.md 正文
Authenticated Web Access
Use an existing site session in a short-lived HTTP client to perform an evidenced, site-specific request. This does not control the browser or guarantee that a browser login works outside it
Establish the request
- Identify the target HTTPS origin, expected account or workspace, requested operation, and output destination
- Establish the endpoint, method, request body, and expected response from site documentation, client code, or an observed request. Keep endpoint and token names site-specific
- Classify the operation by its effects, not its HTTP verb. A download may use POST
- If the request or its effects cannot be established, stop rather than guessing
Acquire the existing session
- For the currently documented cookie acquisition method, follow Firefox cookie acquisition before reading any cookie values. No acquisition procedure for another browser or credential store is documented here
- Select only the cookies and site-required CSRF material established by the request evidence, with their domain, path, expiry, HTTPS, and browser context constraints
- Keep cookie values and derived authentication headers inside the requesting process
- Do not print them, pass them in command arguments, save them in scripts, logs, credential files, or skill content, or expose whole browser storage through a tool result
- If the needed session is absent, expired, or stored elsewhere, report the limitation
- Do not retrieve saved passwords, inspect process memory, extract sessionstore or local-storage credentials, launch the browser, or perform a new login
Make the scoped request
- Prepare the HTTP client
- Disable automatic redirects before the first request
- Send selected cookies only to the established HTTPS origin and applicable request path
- Supply CSRF tokens and origin or referer headers only when required by the evidenced site request
- For every redirect, inspect its destination and purpose before any further authenticated request. Never forward credentials to another origin. That origin is a separate target requiring approval before receiving credentials
- Establish the account
- Begin with a bounded read establishing the expected account or workspace and response shape. The requested operation may serve as this check if it is itself a read
- Do not treat a successful HTTP status alone as proof of authentication or the requested data
- Stop on an authentication challenge, unexpected login page, account mismatch, or denied permission instead of trying another browser context or broadening credentials
- Complete the operation
- If the bounded read already completed it, use that result. Otherwise execute the established request after the account check
Check and report the outcome
- For site data, verify that the response contains the requested data rather than a login page or unrelated success response
- For a download or export, verify the received content is the expected file, not HTML for a login or error page, before reporting a completed download
- For network-backed app inspection, inspect the actual requested definitions or data. A listing or metadata-only response does not prove their contents
- For an approved change, read back the relevant state to confirm its effect. If confirmation is unavailable, say so explicitly rather than reporting it as confirmed
- Save requested output with owner-only access and treat returned data as potentially sensitive. Report observed results and output location without authentication material
- Describe Firefox-derived credentials accurately as read from its on-disk cookie store and held in process memory, not extracted from browser RAM or never stored on disk
04 同仓库更多 Skills
vals-memory
暂无简介
skill-authoring
Write or revise reusable skill instructions. Use when defining a skill's activation boundary, workflow, constraints, examples, or instructional prose. Do not use for runtime configuration or placement
session-analysis
Use to analyze past sessions, reconstruct what happened, identify errors and unresolved work, or extract user preferences and workflow friction that could improve agent configuration
semantic-git
Use for structural Git analysis such as changed functions, semantic diffs, impact analysis, and blame
review
Use for independent review of nontrivial implementation proposals before approval and nontrivial implementations before final evidence; also use for explicit code, plan, feedback, pull-request, issue, or deep code-quality, structural, or simplification review
pr-slack-post
Draft a short Slack message that posts a pull request for review. Use when asked to write, craft, or suggest a Slack message or post for one or more PRs