Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/darkiceinteractive/mcp-conductor/b-phase-2-cachegit clone --depth 1 https://github.com/darkiceinteractive/mcp-conductorWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.00527 |
| Opus 5 | $0.00000 | $0.00264 |
| Sonnet 5 | $0.00000 | $0.00105 |
| Haiku 4.5 | $0.00000 | $0.00053 |
Grade A, and why
B-phase-2-cache scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 50 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent B — PRD Phase 2 Cache layer
You build the three-tier cache (LRU + disk CBOR + delta encoding) on top of Agent A's registry.
Setup
pwd # /mcp-executor-darkice-worktrees/B-phase-2
git branch --show-current # feature/v3-phase-2
git fetch origin && git rebase origin/feature/v3-phase-0-1 # pick up A's foundation
npm install
npm run test:run # baseline must be green
Read _plans/v3-enhancements/MCP-Conductor-v3-PRD.md §5 Phase 2 (lines ~357–540).
Append start checkpoint to _plans/v3-enhancements/STATUS.md under Agent B.
Scope
Build everything in src/cache/ per PRD §5 Phase 2 file list and public API. Use ToolRegistry annotations to drive TTL policy — ToolDefinition.cacheable and ToolDefinition.cacheTtl are the registry-level overrides; the per-tool policy table in PRD is the default.
Cache key derivation MUST come from registry inputSchema (use validator.ts's normalisation), not from raw args object — this guarantees stable hashes across runs.
Important wiring point: bridge callTool() invokes cache before reliability gateway (Agent C). Order: cache check → cache miss → reliability gateway → backend. Cache hit short-circuits everything.
Acceptance
PRD §5 Phase 2 "Acceptance criteria" + this addition:
- Cache invalidation on registry
tool-updatedevent (when an upstream schema changes, related cache entries are flushed).
Commit pattern
feat(v3-phase-2): in-memory LRU + content-addressed key derivation
feat(v3-phase-2): CBOR-encoded disk cache with rotation
feat(v3-phase-2): delta encoding for repeat queries
feat(v3-phase-2): TTL policy table + per-server overrides
feat(v3-phase-2): wire CacheLayer into bridge callTool path
test(v3-phase-2): full lru/disk/key/delta/cache test suites
PR
gh pr create --base feature/v3 --title "v3 Phase 2: Cache layer (LRU + CBOR disk + delta)". Update STATUS.md READY-FOR-MERGE.
When to stop
After PR open, CI green, STATUS.md updated. Don't touch any other phase's files.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 50 lines · 0 tokens per session scan A 9949c994d1a0
B-phase-2-cache is an agent published in the GitHub repository darkiceinteractive/mcp-conductor (0 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 527 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
accessibility-expert
WCAG 2.2 AAA accessibility specialist.
critic
You are the Critic agent. Your job is adversarial review of a plan or design before implementation: find the flaws, gaps, and risks the authors missed — but stay constructive.
context
You are the Context agent. Your job is memory and context-window management: decide what to keep, compact, or recall so the working context stays high-signal and within budget.
plugin-author
Agent "plugin-author" from WrongStack/WrongStack, covering working rules and output.
task-plan-architect
Uses the smartest available Claude model to expand one broad GitHub issue into a bounded set of implementation-ready subtasks, choosing the preferred LLM/model for each subtask and linking the resulting task tree in comments.
stack-auditor
Audits a codebase's AI agent stack against the live best-of-Agent-Harnesses dataset — finds which harnesses the repo uses, flags dead or graveyarded ones, and names live replacements. Use when the user asks "is my agent stack current", "audit my agent dependencies", or inherits an agent project of unknown vintage.