Dependencies

L1 Improvement

• Reframed the dependency hub using Prompt Architect constraint extraction and Skill Forge structure-first guardrails.
• Added explicit routing rules (mapper vs audit), confidence ceilings, and memory tagging.
• Simplified execution phases to keep graphing, security, and freshness checks coherent.

STANDARD OPERATING PROCEDURE

Purpose

Provide a single entry point for dependency graphing, vulnerability checks, and upgrade planning while routing specialized work to the dependency-mapper subskill when mapping is primary.

Trigger Conditions

• Positive: dependency graph request, vulnerability/audit needs, upgrade planning, circular/unused dependency checks.
• Negative: pure build failures (route to build/debug skills) or single-library research (route to language specialist).

Guardrails

• Structure-first docs: SKILL.md, README.md, QUICK-REFERENCE.md, examples/tests placeholders maintained.
• Use explicit ecosystem detection (npm/pnpm/yarn, pip/poetry, cargo, go mod, maven/gradle, composer); cite evidence.
• Confidence ceilings enforced on findings; call out unverifiable CVEs.
• Memory tagging required for audits and graphs.

Execution Phases

1. Intent & Ecosystem Detection – Identify package managers, lockfiles, and scope; decide if routing to dependency-mapper is needed.
2. Data Collection – Build graph/tree, fetch vulnerability reports, capture outdated/unused/circular signals.
3. Analysis – Summarize risks (CVE, license, freshness) and stability constraints; map blast radius for upgrades.
4. Plan & Propose – Recommend remediation (pin, upgrade, remove, isolate); include commands and expected impacts.
5. Validation – Run audits/tests as available; ensure graph/risk deltas improve.
6. Delivery – Provide findings, proposed actions, and confidence line with ceiling syntax; archive in memory.

Output Format

• Ecosystem detection and scope.
• Graph summary (nodes/edges/risks) and key vulnerabilities/outdated packages.
• Remediation plan with commands, owners, and blast radius notes.
• Confidence: X.XX (ceiling: TYPE Y.YY) and memory keys used.

Validation Checklist

• Ecosystem detected and mapped; routing decision documented.
• Vulnerability/freshness data captured with sources.
• Remediation plan includes rollback/lockfile guidance.
• Memory tags applied and reports stored.
• Confidence ceiling declared.

Integration

• Subskill: use when-mapping-dependencies-use-dependency-mapper for deep graph visualization.
• Memory MCP: skills/tooling/dependencies/{project}/{timestamp} for audits, graphs, and remediation notes.
• Hooks: pre/post targets aligned to Skill Forge latency bounds.

Confidence: 0.70 (ceiling: inference 0.70) – SOP rewritten to align with Prompt Architect and Skill Forge cadence.