← ClaudeAtlas

intent-driftlisted

Detects intent drift in AI-assisted development by comparing the original goal against the current plan and execution context. Run /intent-drift to evaluate whether ongoing work has drifted from what was asked. Triggers on "check intent drift", "are we still on track", "is this aligned with the goal", "did the plan drift", "intent alignment", "scope creep check", "am I off track".
shauryagangrade/intent-drift-skill · ★ 11 · AI & Automation · score 70
Install: claude install-skill shauryagangrade/intent-drift-skill
# intent-drift A skill that uses the Intent Alignment Engine to detect whether an AI coding agent's current work has drifted from the user's original goal. ## Run (when invoked) **0. Gather inputs** (ask only what you cannot infer from the repo): - `--original-goal` (required): the user's stated objective, or infer from the conversation / issue / PR description if available. - `--current-plan` (required): what is actually being built right now (steps, branch purpose, recent edits), or infer from git state + recent messages. - `--context` / `--auto-context`: execution evidence (edited files, git diff, recent commands). With `--auto-context`, collect automatically (see `scripts/collect_context.py`). - `--format` (`text` | `markdown` | `json`): output format. - `--threshold` (default 75): minimum alignment % to be considered "on track". Do NOT ask if enough is inferable from the repo + conversation — infer and state your assumptions. **1. Build the context** and call the analyzer: ```bash python3 ~/.claude/skills/intent-drift/analyzer.py \ --original-goal "Reduce memory usage" \ --current-plan "Optimize startup latency" \ --context "Edited: main.py, startup.py. Recent: pip install numpy, cProfile main.py" \ --format text ``` Or in Python: ```python sys.path.insert(0, str(Path.home() / ".claude" / "skills" / "intent-drift")) from analyzer import IntentDriftAnalyzer a = IntentDriftAnalyzer() rep = a.analyze(a.parse_arguments([ "--original-goal", "Reduce memo