Back to Skills

shift-camouflage

pjt222
Updated Yesterday
3 views
17
2
17
View on GitHub
Developmentapi

About

This skill enables adaptive, polymorphic APIs that present different interfaces based on environment context, similar to a cuttlefish changing its appearance. It reduces attack surfaces and enables feature flagging by dynamically altering the exposed surface layer without changing the core logic. Use it for context-aware behavior, progressive rollouts, and obscuring system patterns from different observers.

Quick Install

Claude Code

Recommended
Primary
npx skills add pjt222/agent-almanac -a claude-code
Plugin CommandAlternative
/plugin add https://github.com/pjt222/agent-almanac
Git CloneAlternative
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/shift-camouflage

Copy and paste this command in Claude Code to install this skill

Documentation

Shift Camouflage

Adaptive surface transform — polymorphic interfaces, context-aware behavior, dynamic presentation. Cuttlefish chromatophores. Surface adapts → env, core stable. Reduces attack surface + optimizes diverse observer interaction.

Use When

  • Diff interfaces → diff consumers (API ver, multi-tenant, role-based)
  • Reduce attack surface → expose only what observer needs
  • Feature flags, progressive rollout, A/B at interface
  • Adapt behavior → env context w/o core change
  • Protect internal arch from external coupling (observers couple surface, not structure)
  • Complement adapt-architecture when surface enough, deep transform unneeded

In

  • Required: System whose surface adapts
  • Required: Observers + diff interface needs
  • Optional: Current interface design + limits
  • Optional: Threat model (hide what from whom?)
  • Optional: Feature flag | progressive rollout infra
  • Optional: Perf constraints (dynamic surface gen has overhead)

Do

Step 1: Map Observer Landscape

Who interacts + what each needs to see.

  1. Catalog observers:
    • External (end users, API consumers, partners)
    • Internal services (microservices, bg jobs, admin tools)
    • Adversaries (attackers, scrapers, competitors)
    • Regulators (auditors, compliance)
  2. Per observer:
    • Need to see (req surface)
    • Should not see (hidden)
    • Expect to see (compat surface — may differ from need)
    • How interact (protocol, freq, sensitivity)
  3. Build observer-surface matrix:
Observer-Surface Matrix:
┌──────────────┬────────────────────────┬─────────────────┬──────────────┐
│ Observer     │ Required Surface       │ Hidden Surface  │ Threat Level │
├──────────────┼────────────────────────┼─────────────────┼──────────────┤
│ End users    │ Public API v2, UI      │ Internal APIs,  │ Low          │
│              │                        │ admin endpoints │              │
├──────────────┼────────────────────────┼─────────────────┼──────────────┤
│ Partner API  │ Partner API, webhooks  │ Internal logic, │ Medium       │
│              │                        │ user data       │              │
├──────────────┼────────────────────────┼─────────────────┼──────────────┤
│ Admin tools  │ Full API, debug        │ Raw data store  │ Low          │
│              │ endpoints              │ access          │              │
├──────────────┼────────────────────────┼─────────────────┼──────────────┤
│ Adversaries  │ Nothing (minimal)      │ Everything      │ High         │
│              │                        │ possible        │              │
└──────────────┴────────────────────────┴─────────────────┴──────────────┘

Got: Complete observer landscape w/ surface reqs. Drives all camouflage design.

If err: Incomplete obs ID → start two extremes (most privileged: admin; most restricted: adversary). Design surfaces, interpolate between.

Step 2: Design Chromatophore Mapping

Map observer context → surface presentation. "Chromatophore" layer.

  1. Context signals:
    • Auth identity → privilege
    • Origin → geo, network, app
    • Feature flags → enable/disable
    • Time/phase → deploy stage, biz hours, maint
    • Load/health → degraded mode → reduced surface
  2. Surface gen rules. Per context combo, elements are:
    • Visible: in res/interface
    • Hidden: excluded entirely (errs reveal nothing)
    • Transformed: present but modified for observer (diff schema, simpler data)
    • Decoy: deliberately misleading for adversarial contexts
  3. Implement chromatophore layer:
    • Thin middleware/proxy between core + observers
    • Eval context signals each req
    • Apply surface config
    • Never modify core behavior — only filter + transform surface
Chromatophore Architecture:
┌──────────────────────────────────────────────────────┐
│ Observer Request                                      │
│        │                                              │
│        ↓                                              │
│ ┌─────────────────┐                                   │
│ │ Context Extract  │ ← Auth, origin, flags, time      │
│ └────────┬────────┘                                   │
│          ↓                                            │
│ ┌─────────────────┐                                   │
│ │ Surface Select   │ ← Observer-surface matrix lookup  │
│ └────────┬────────┘                                   │
│          ↓                                            │
│ ┌─────────────────┐                                   │
│ │ Core System      │ ← Processes request normally      │
│ └────────┬────────┘                                   │
│          ↓                                            │
│ ┌─────────────────┐                                   │
│ │ Surface Filter   │ ← Remove/transform/add elements   │
│ └────────┬────────┘                                   │
│          ↓                                            │
│ Observer Response (adapted surface)                    │
└──────────────────────────────────────────────────────┘

Got: Mapping translates observer context → surface config. Explicit, auditable, separate from core.

If err: Too complex → simplify to role-based: 3-5 profiles (public, partner, admin, internal, minimal). Map every observer → one.

Step 3: Behavioral Polymorphism

Behavior adapts to context, not just surface.

  1. Context-dep behaviors:
    • Res detail (verbose admin, minimal public)
    • Rate limit (generous partners, strict unknown)
    • Err msgs (detail internal, generic external)
    • Data freshness (real-time premium, cached std)
    • Feature avail (full beta, stable-only general)
  2. Variants:
    • Each = complete tested path
    • Context → which variant runs
    • Variants share core, differ in presentation + policy
  3. Feature flag integration:
    • Flags control active variants
    • Progressive rollout: % of observers, increase over time
    • Circuit breakers: auto-revert safe behavior on err

Got: Behavior adapts → context. Same core → appropriate res for diff audiences. Flags → progressive rollout.

If err: Too many code paths → consolidate pipeline: core → policy layer → presentation layer. Polymorphism in policy + presentation only, core singular.

Step 4: Reduce Attack Surface

Minimize what adversaries observe + interact w/.

  1. Least surface:
    • Each observer sees only what needed
    • Unauth observers see min possible
    • Errs never leak internals (no stack traces, paths, vers)
  2. Active reduction:
    • Remove default pages, headers, endpoints revealing tech stack
    • Randomize non-essential res chars (timing jitter, header order)
    • Disable unused endpoints entirely (off, not hidden)
  3. Pattern disruption:
    • Vary res chars → defeat fingerprint
    • Controlled unpredictability in non-functional aspects
    • Functional behavior deterministic, surface chars vary
  4. Recon monitoring:
    • Detect req patterns probing hidden surface (enum attacks)
    • Alert repeated access to nonexistent endpoints (path fuzz)
    • Track + correlate recon across sessions (see defend-colony)

Got: Min attack surface. Adversaries can't ID stack, internals, hidden caps. Recon detected + tracked.

If err: Reduction breaks legit consumers → matrix incomplete. Review Step 1, update. Randomization issues → reduce to non-functional only (timing, headers), keep functional res deterministic.

Step 5: Surface Coherence

Dynamic surface stays consistent, debuggable, maintainable.

  1. Testing:
    • Each profile explicit (admin sees admin? public sees public?)
    • Transitions (context changes mid-session?)
    • Failure modes (chromatophore layer fails → what surface?)
  2. Docs:
    • Each profile + config
    • Context signals + effects
    • Sync w/ actual behavior (test docs vs reality)
  3. Debug:
    • Admin/debug mode → which profile active + why
    • Logs → which config applied per req
    • Replay req through specific profile
  4. Evolution:
    • Add: appropriate profiles, test, deploy
    • Remove: deprecation warning, then remove
    • Change: flag controlled, progressive rollout

Got: Maintainable, testable, documented system. Dynamic ≠ undebuggable.

If err: Debug nightmare → add transparency: trace header (admin/debug only) → which profile applied + which signals decided.

Check

  • Observer landscape mapped w/ surface reqs
  • Chromatophore translates context → surface config
  • Behavioral polymorphism adapts to context
  • Attack surface min for adversaries
  • Each profile explicit tested
  • Failure mode → safe default (minimal)
  • Debug/admin can inspect active config
  • Docs match behavior

Traps

  • Complexity explosion: Too many profiles + variations. Max 3-5 profiles.
  • Core contamination: Surface logic leaks into core. Chromatophore = separate. If-statements about observer type in core code → arch wrong.
  • Obscurity alone: Surface reduction = defense-in-depth, not auth/authz replacement. Hidden endpoint still needs authn+authz.
  • Inconsistent surfaces: A sees v1, B sees v2, supposed same. Test explicit, matrix authoritative.
  • Failure surface: Chromatophore fails → what does observer see? Default must be safe (minimal), not open (full).

  • assess-form — surface adaptation may resolve form pressure w/o deep transform
  • adapt-architecture — deep structural change when surface insufficient
  • repair-damage — surface can mask damage during repair (caution — don't hide real probs)
  • defend-colony — attack surface reduction = defense layer
  • coordinate-swarm — context-aware in distributed needs coordinated surface
  • configure-api-gateway — API gateways implement chromatophore in practice
  • deploy-to-kubernetes — k8s svc + ingress enable network-level surface control

GitHub Repository

pjt222/agent-almanac
Path: i18n/caveman-ultra/skills/shift-camouflage
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

Related Skills

qmd

Development

qmd is a local search and indexing CLI tool that enables developers to index and search through local files using hybrid search combining BM25, vector embeddings, and reranking. It supports both command-line usage and MCP (Model Context Protocol) mode for integration with Claude. The tool uses Ollama for embeddings and stores indexes locally, making it ideal for searching documentation or codebases directly from the terminal.

View skill

subagent-driven-development

Development

This skill executes implementation plans by dispatching a fresh subagent for each independent task, with code review between tasks. It enables fast iteration while maintaining quality gates through this review process. Use it when working on mostly independent tasks within the same session to ensure continuous progress with built-in quality checks.

View skill

mcporter

Development

The mcporter skill enables developers to manage and call Model Context Protocol (MCP) servers directly from Claude. It provides commands to list available servers, call their tools with arguments, and handle authentication and daemon lifecycle. Use this skill for integrating and testing MCP server functionality in your development workflow.

View skill

adk-deployment-specialist

Development

This skill deploys and orchestrates Vertex AI ADK agents using A2A protocol, managing AgentCard discovery, task submission, and supporting tools like Code Execution Sandbox and Memory Bank. It enables building multi-agent systems with sequential, parallel, or loop orchestration patterns in Python, Java, or Go. Use it when asked to deploy ADK agents or orchestrate agent workflows on Google Cloud.

View skill