Pluribus Context Receipts
Privacy-safe evidence for AI context boundaries: what crossed into the agent, what stayed out, and what should stop before edits or tool calls.
30-second context sync demo
Edit one canonical project rule, preview the native CLAUDE.md, AGENTS.md, or .cursorrules, and inspect a source→output receipt. It runs entirely in your browser and keeps the honest boundary: generated does not mean loaded.
Field evidence
Inspect boundary checks that independent maintainers merged into skill, authority-snapshot, context-freshness, rollback, handoff, and instruction-integrity projects. The ledger separates accepted implementation evidence from adoption claims.
Boundary receipt gallery
If you are comparing Pluribus with memory servers, MCP gateways, AGENTS.md/CLAUDE.md frameworks, Skills, or multi-session orchestration tools, start with the gallery. It groups the public examples by the boundary they prove: tool identity, stale rule authority, instruction load safety, long-session resume state, parallel-session ownership, and memory/RAG authority.
Browser playground
Try receipt validation without installing anything or sending data to a server. Current samples cover MCP tool-surface diffs, retrieved-context attention, CLAUDE.md rule-attention drift, skill install provenance, and retrieval adoption evidence. If you are debugging MCP context bloat before startup, use the task-scoped MCP config receipt demo. If a Skill or paste-cleaning CLI claims token savings, use the semantic anchor preservation receipt demo to prove it kept version/API/security anchors. If a Cursor/Claude Code workflow requires a first tool or pre-tool hook before edits, use the session preflight receipt and copyable rule/example. If a memory backend, GraphRAG layer, or handover skill returns useful facts, use the memory provenance + authority-home receipt and tryable example to prove which memory was used, suppressed, or rejected as stale. If a config doctor rewrites scattered AI-agent rules, use the config treatment receipt to prove the canonical authority loaded in each target surface without copying private rule payloads. If archived agent chats disappear under Skill/search noise, load the archived chat recovery card sample in the playground to check whether a result is actually resumable. If a HANDOFF.md/CONTEXT.md-style skill claims session continuity, load the handoff freshness receipt sample to prove the handoff still matches the source session, repo state, checks, and referenced files before a new agent resumes. If context/memory files are starting to bloat every new session, load the context diet sample to prove a note was promoted only after repeated observed corrections, scoped to a task/path, and paired with pruning evidence. If a `/hygiene`, `/doctor`, memory-kit, or MCP cleanup wants to delete/disable context, load the context hygiene sample first to audit loaded sources, candidate removals, safety negative controls, rollback, and review gates before pruning.
CLI quick checks
npx --yes pluribus-context@latest demo tool-surface-diff --json
npx --yes pluribus-context@latest validate
Agent Skills discovery
Pluribus publishes a well-known Agent Skills index so compatible clients can discover the low-authority context-receipts and skill-policy-receipts skills without scraping GitHub.