{"id":"sheeki03-few-word","name":"Few-Word","homepage":"https://sheeki03.github.io/Few-Word/","repo_url":"https://github.com/sheeki03/Few-Word","category":"ai-ml","subcategories":[],"tags":["ai-agents","claude-code","plugins","context-engineering","token-optimization","filesystem-memory","debugging"],"what_it_does":"FewWord is a Claude Code plugin that intercepts large shell command outputs and offloads them to files on disk, replacing them in-context with compact pointers (plus small previews for failures). It also provides commands to open, search, and manage offloaded outputs and persists/archives plans via hook events.","use_cases":["Reduce token/context bloat from large command outputs (tests, logs, build artifacts)","Debugging: quickly retrieve failure context (tail/preview plus full output on demand)","Keeping agent plans and step outputs retrievable across context compaction/summarization","Organizing session artifacts with retention/cleanup policies (TTL + LRU)"],"not_for":["Commands with interactive or full-screen behavior (ssh, vim, less, top, etc.)","Commands where output is already redirected or where pipelines/heredocs are used (explicitly limited in v1 per README)","Use cases requiring guaranteed end-to-end data portability/security guarantees beyond local filesystem handling"],"best_when":"You frequently run commands that produce large stdout/stderr and you want the agent to stay within context limits while still being able to retrieve full outputs when needed.","avoid_when":"You cannot tolerate local disk writes of command outputs (even with redaction) or you rely heavily on unsupported shell constructs (pipelines/heredocs) that FewWord skips or cannot offload in v1.","alternatives":["Manual output redirection to files (e.g., tee > artifacts.txt) and prompting the model to reference those files","Other context-engineering/file-based retrieval approaches (e.g., custom hooks + retrieval tooling)","LangChain/file-system based context strategies (as referenced by the project)"],"af_score":61.5,"security_score":33.0,"reliability_score":36.2,"package_type":"skill","discovery_source":["openclaw"],"priority":"high","status":"evaluated","version_evaluated":null,"last_evaluated":"2026-03-30T15:27:22.022463+00:00","interface":{"has_rest_api":false,"has_graphql":false,"has_grpc":false,"has_mcp_server":false,"mcp_server_url":null,"has_sdk":false,"sdk_languages":["python"],"openapi_spec_url":null,"webhooks":false},"auth":{"methods":["None mentioned (plugin installation via Claude plugin marketplace commands)"],"oauth":false,"scopes":false,"notes":"README describes environment variables for behavior/escape hatches but no user auth flows for a network API."},"pricing":{"model":null,"free_tier_exists":false,"free_tier_limits":null,"paid_tiers":[],"requires_credit_card":false,"estimated_workload_costs":null,"notes":"Pricing not described in provided README content."},"requirements":{"requires_signup":false,"requires_credit_card":false,"domain_verification":false,"data_residency":[],"compliance":[],"min_contract":null},"agent_readiness":{"af_score":61.5,"security_score":33.0,"reliability_score":36.2,"mcp_server_quality":0.0,"documentation_accuracy":75.0,"error_message_quality":0.0,"error_message_notes":null,"auth_complexity":100.0,"rate_limit_clarity":60.0,"tls_enforcement":0.0,"auth_strength":40.0,"scope_granularity":0.0,"dependency_hygiene":40.0,"secret_handling":85.0,"security_notes":"README claims HTTPS/TLS is not applicable (local plugin), but describes several security hardening measures: path traversal prevention, bounded file reads (skipping reads >2MB), secret redaction (AWS keys, GitHub tokens, API keys), and manifest integrity via robust JSON escaping. It also notes that raw command arguments/tokens/secrets may be present in logs or not logged depending on category, and MCP tool calls clamp pagination and avoid logging argument values by logging input parameter keys/counts instead.","uptime_documented":0.0,"version_stability":55.0,"breaking_changes_history":30.0,"error_recovery":60.0,"idempotency_support":"false","idempotency_notes":null,"pagination_style":"none","retry_guidance_documented":false,"known_agent_gotchas":["FewWord skips certain command types (interactive commands, heredocs, and pipelines in v1) which may lead to large outputs remaining in-context.","Automatic behavior can be disabled via escape hatch files/env vars; agents should account for different plugin modes.","Retention/cleanup (TTL + LRU) means previously offloaded outputs might be unavailable after cleanup unless pinned."]}}