Aidlc Tracker

AI-DLC Tracker

A lightweight status/reminder layer on top of the AI-DLC methodology (Inception → Construction → Operations, with an optional Verification phase for assessment-style projects). This skill tracks where the project is and what's next — it does not replace a fuller AI-DLC rules framework if one is checked into the repo (see "Defer to a fuller framework" below).

1. Find the tracking docs

Check, in this order, for the AI-DLC docs root:

• docs/aidlc/ (common when a project remaps the official path)
• aidlc-docs/ (the official AI-DLC default)
• .aidlc-docs/

Within that root, the key files are:

• aidlc-state.md — current phase/stage/status + phase-completion table
• audit.md — append-only log of stage approvals and decisions

If none of these exist, this project hasn't been initialized for AI-DLC tracking — see "Bootstrapping a new project" below instead of skipping ahead.

2. Report current status

Read aidlc-state.md and summarize for the user:

• Current phase, current stage, status, blockers, last-updated date
• The Phase Completion table (which phases are ✅ Complete / 🔄 In Progress / ⬜ Not Started)
• Any non-default Extension Configuration entries

3. Determine what's next

Use this condensed phase/stage map (full detail lives in a fuller framework if present — see below):

| Phase | Stages (✅ = always, ⬪ = conditional) | Typical artifacts | |---|---|---| | Inception | ✅ Workspace Detection → ⬪ Reverse Engineering (brownfield) → ✅ Requirements Analysis → ⬪ User Stories → ✅ Workflow Planning → ⬪ Application Design → ⬪ Units Generation | inception/architecture-decisions.md, inception/execution-plan.md, inception/requirements-*.md | | Construction (per unit) | ⬪ Functional Design → ⬪ NFR Requirements → ⬪ NFR Design → ⬪ Infrastructure Design → ✅ Code Generation, then ✅ Build and Test (after all units) | construction/functional-design.md, construction/nfr-*.md, construction/infrastructure-design.md, construction/build-and-test/*.md | | Operations | Deployment, monitoring, runbooks (project-defined) | operations/deployment-runbook.md, operations/deployment-readiness-checklist.md, operations/e2e-testing-guide.md | | Verification (if used) | Self-check / readiness review, interview prep | verification/self-check-report.md, verification/interview-prep.md |

To propose "what's next":

1. Cross-reference the Phase Completion table against which artifact files

actually exist (Glob the phase folders) — a phase marked complete with missing expected artifacts is worth flagging.

1. Identify the first phase/stage that is not yet ✅ Complete.
2. Propose that as the next step, but do not start it without confirming

— see hard rules below.

4. Hard rules (apply regardless of which framework is present)

• Never silently advance phases. Each stage ends with an explicit

approval question to the user. Do not start the next stage until they confirm.

• Audit log is append-only. Never overwrite audit.md. Add a new row/entry

with a timestamp, the phase/stage, a short description of what happened, and who acted (User/Agent). Match whatever format already exists in the file — if it's a markdown table, append a row; if it's per-entry blocks, append a block. Don't reformat existing entries.

• Update aidlc-state.md in the same turn work completes — bump the

Phase Completion table, Current Status (phase/stage/last-updated), and Blockers. Stale state files are worse than none.

• Construction per-unit stages use a 2-option completion message

("Request Changes" vs. "Continue to Next Stage") — don't invent a 3rd option or skip the gate.

• Plan checkboxes: if a stage produced a plan file with - [ ] items,

check them off - [x] immediately as each item completes, in the same turn — not in a later cleanup pass.

5. Defer to a fuller framework if present

If the repo also contains a full AI-DLC rules framework — look for AIDLC/aidlc-rules/, .aidlc-rule-details/, .kiro/aidlc-docs/, or .amazonq/aws-aidlc-rule-details/ — that framework's per-stage rule files are authoritative for how to execute a stage (required questions, content validation, message formats). Use this skill for the status/tracking layer (steps 1-4) and load the matching rule file from that framework before executing a specific stage's detailed steps.

6. Bootstrapping a new project

If no AI-DLC docs exist yet and the user wants to start tracking:

1. Create docs/aidlc/aidlc-state.md with: a Project table (scenario, repo,

doc root), a Current Status table (phase=Inception, stage=Workspace Detection, status, last-updated=today), an empty Phase Completion table (Inception/Construction/Operations all ⬜ Not Started), and an Artifacts Index pointing at the phase subfolders.

1. Create docs/aidlc/audit.md with a header and the table format shown

above (| Timestamp (UTC) | Phase / Stage | Action | User / Agent |).

1. Create empty inception/, construction/, operations/ subfolders as

work begins (don't pre-create empty folders speculatively — create each when its first artifact is written).

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: sergeyitaly
• Source: sergeyitaly/claude-skill-deployer
• License: MIT

Install and usage instructions live in the source repository linked above.