More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Decision Trail: Why Your Project Is the Way It Is

    1

    You get back a dated timeline of the decisions that shaped the project, each with its reason and the exact place in the sources that backs it.

    $15

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Decision Trail: Why Your Project Is the Way It Is

    Decision Trail: Why Your Project Is the Way It Is

    Example session with this skill installed

    I joined last week. Build a decision timeline from the commits, the ADRs and the meeting notes, and tell me which decisions have no recorded reason.

    • Read your context and instructions
    • Compiled the decision trail

    trail_check.py:
    ADVICE [undated] E074 has no date in its source and is cited by D05: say so where it is used
    summary: 0 FIX, 1 ADVICE in en-trail (74 events, 10 decisions, 13 reasons: 11 stated, 1 inferred, 1 unknown; 3 threads). This checks evidence links, not whether the story is complete: history that was never written down is not in the trail.

    From the trail

    • D06 Move from SQLite to PostgreSQL (16 June to 8 July 2025), replaces D01. Reason: stated, quoted from ADR-0002.
    • D05 Remove the PDF export (28 May 2025). Reason: inferred from the timeout ticket and two repair commits; no source states it.
    • D08 Shorten sessions from 30 days to 12 hours (17 September 2025). Reason: unknown. Open question for the team.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    English

    For developers and team leads who join a project and ask: why is it like this?

    Give your agent the project's history: a git log, ticket exports, decision records, meeting notes. You get back a dated timeline of the decisions that shaped the project, each with its reason and the exact place in the sources that backs it.

    What you get each run

    • timeline.html: one printable page. The events in date order, the decisions standing out, and the threads that connect them (for example "Where the data lives").
    • decisions.md: the same decisions as Markdown, for a wiki or an onboarding handbook.
    • decisions.json: each decision with what was decided, when, why, who, the alternatives, what it replaced and what was seen later.
    • events.json: every dated event, pointing at its exact place in the raw source.

    Every reason is labelled

    • Stated: a word-for-word quote from a source, with the event it came from.
    • Inferred: the agent's own reading of the timing and the evidence, said plainly as a reading.
    • Unknown: no source gives a reason. It is carried as an open question for the team.

    Evidence, checked by script

    A bundled Python script (standard library, read-only, no network calls) builds the event list from the raw files. A checker then verifies every quote letter for letter against its source and every link between a decision and its events. The agent fixes and checks again until no FIX is left.

    What it reads

    A git log export (the agent runs git; the scripts never do), ticket exports as CSV or JSON (Jira, Linear, GitLab, GitHub issues), ADR or RFC Markdown files, meeting notes, and chat exports as text. Author email addresses are left out of the git export on purpose.

    Honest about its limits

    A trail holds only what was written down. History that never reached a commit, a ticket or a note is not in it, and the trail lists what the sources do not cover. The checker verifies evidence links, not whether the story is complete. Decisions are described without blame, and people appear only as the sources name them.

    Ask it questions afterwards

    "Why did we drop the PDF export?" The agent answers from the trail with the decision, its date and its events, and says so when the sources give no reason.

    Arabic

    The timeline and the decisions can be written in Arabic with a right-to-left page. Quotes stay in the language of their source, and hashes and file names stay as they are.

    In the download: SKILL.md, the scripts, references (sources, the decisions format, reading history, Arabic notes), and a complete example so you can see the output before you buy: fourteen months of a fictional project (53 commits, 11 tickets, 3 decision records, meeting notes) and the English and Arabic trails made from it.

    بالعربي

    للمبرمجين وقادة الفرق الذين ينضمون إلى مشروع ويسألون: لماذا هو هكذا؟

    أعطِ وكيلك تاريخ المشروع: سجل git، وتصدير التذاكر، وسجلات القرار، ومحاضر الاجتماعات. تستلم خطًا زمنيًا مؤرَّخًا للقرارات التي صنعت المشروع، وكل قرار معه سببه والموضع الدقيق في المصادر الذي يسنده.

    ماذا تستلم في كل مرة

    • صفحة timeline.html: صفحة واحدة قابلة للطباعة فيها الأحداث بترتيب التاريخ، والقرارات بارزة، والخيوط التي تربط بينها.
    • ملف decisions.md: القرارات نفسها بصيغة Markdown، لويكي الفريق أو دليل المنضمين الجدد.
    • ملف decisions.json: كل قرار مع ما تقرر، ومتى، ولماذا، ومن، والبدائل، وما الذي حلّ محله، وما ظهر لاحقًا.
    • ملف events.json: كل حدث مؤرَّخ مع موضعه الدقيق في المصدر الخام.

    كل سبب له وصف

    • مذكور: اقتباس حرفي من مصدر، مع الحدث الذي جاء منه.
    • مستنتج: قراءة الوكيل للتوقيت والأدلة، ومكتوب بوضوح أنها قراءة.
    • غير معروف: لا يذكر أي مصدر سببًا، فيبقى سؤالًا مفتوحًا للفريق.

    أدلة يدققها سكربت

    سكربت بايثون مرفق (يقرأ فقط، بلا اتصال بالشبكة) يبني قائمة الأحداث من الملفات الخام. ثم أداة فحص تطابق كل اقتباس حرفًا بحرف مع مصدره وكل رابط بين قرار وأحداثه. يصلح الوكيل ويعيد الفحص حتى لا يبقى أي FIX.

    ماذا يقرأ

    تصدير سجل git (الوكيل يشغّل git، والسكربتات لا تشغّله)، وتصدير التذاكر بصيغة CSV أو JSON، وملفات ADR أو RFC، ومحاضر الاجتماعات، وتصدير المحادثات كنص. وعناوين بريد المؤلفين تُترك خارج تصدير git عمدًا.

    صادق في حدوده

    السجل يحمل ما كُتب فقط. التاريخ الذي لم يصل إلى إيداع أو تذكرة أو محضر ليس فيه، والسجل يذكر ما لا تغطيه المصادر. وأداة الفحص تتحقق من روابط الأدلة، لا من اكتمال القصة. والقرارات تُوصف بلا لوم لأحد.

    اسأله بعد ذلك

    «لماذا تخلّينا عن تصدير PDF؟» يجيب الوكيل من السجل بالقرار وتاريخه وأحداثه، ويقول ذلك صراحة إذا لم تذكر المصادر سببًا.

    العربية

    يُكتب الخط الزمني والقرارات بالعربي في صفحة من اليمين إلى اليسار. الاقتباسات تبقى بلغة مصدرها، وأسماء الملفات كما هي.

    داخل الملف: ملف SKILL.md، والسكربتات، والمراجع، ومثال كامل لترى النتيجة قبل الشراء: أربعة عشر شهرًا من مشروع خيالي (53 إيداعًا، و11 تذكرة، و3 سجلات قرار، ومحاضر اجتماعات) والسجلان الإنجليزي والعربي المصنوعان منه.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Technical Documentation.

    A trail holds only what was written down: history that never reached a commit, a ticket or a note is not in it, and it lists what the sources do not cover. The checker verifies quotes and evidence links, not whether the story is complete. The scripts read text exports (git log, CSV or JSON tickets, Markdown records and notes, text chat exports); they do not connect to Jira, Linear, GitHub or Slack, so you export the files yourself. Number-only dates such as 03/04/2025 are treated as undated. timeline.html loads its fonts from Google Fonts when online and falls back to system fonts offline. السجل يحمل ما كُتب فقط، والأداة لا تتحقق من اكتمال القصة.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 3 days ago

    Software
    PythonGit

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    Reads the source files you put in trail/sources, writes events.json, decisions.json, timeline.html and decisions.md into the trail folder, and runs its local Python scripts (trail_build.py, make_timeline.py, trail_check.py). The agent runs one git log command to export the history; the scripts run no commands and make no network calls. يقرأ ملفات المصادر ويكتب ملفات السجل فقط.

    Listed3 days ago

    What's inside

    Frequently Asked Questions

    Text exports of the project's history: a git log (the agent exports it), tickets as CSV or JSON, ADR files, meeting notes, chat exports. Take what exists; it works with only some of them. تصدير نصي لتاريخ المشروع، ويعمل بما هو موجود منه.

    No. A stated reason is a word-for-word quote that the checker verifies against the source. A reading is labelled inferred, and a missing reason is labelled unknown. لا، السبب المذكور اقتباس حرفي تتحقق منه أداة الفحص.

    No. The scripts read files and make no network calls. You export the tickets or chats yourself and put them in the sources folder. لا، يقرأ ملفات تصدّرها أنت.

    No. Decisions are described, not people's competence, and people appear only as the sources name them. Author email addresses are left out of the git export. لا، يصف القرارات لا الأشخاص.

    No, and it never says so. It holds what was written down and lists what the sources do not cover. لا، يحمل ما كُتب فقط ويذكر ما لا تغطيه المصادر.

    Yes. Add a new export or a dated note and build again; event ids are kept, so existing citations still point at the right place. نعم، أضف مصدرًا جديدًا وأعد البناء.

    Yes, as an added option: an Arabic timeline with a right-to-left page. Quotes stay in the language of their source. نعم، والاقتباسات تبقى بلغة مصدرها.

    New in Technical Documentation