{ "checker": { "claims": [ { "claim": "Archcore writes plain Markdown documents into .archcore/ and the MCP server returns them", "command": " mcp --project tools/call {\"tool\": \"get_document\", \"arguments\": {\"path\": \".archcore/use-redis-token-bucket.adr.md\"}}", "evidence": "Use Redis for the token bucket\\\",\\\"status\\\":\\\"draft\\\",\\\"mtime\\\":\\\"2026-10-25T23:05:04.800758815Z\\\",\\\"content\\\":\\\"---\\\\ntitle: \\\\\\\"Use Redis for the token bucket\\\\\\\"\\\\nstatus: draft\\\\n---\\\\n\\\\n## Context\\\\n\\\\nState the trigger in 2 to 4 sentences: the situation now, the constraint or\\\\nmeasurement that turns it into a problem, and what breaks if nothing changes.\\\\nCite the evidence \u2014 @path/to/file, a metric, a commit, an external authority \u2014\\\\nor mark the claim [assumption] when the decision runs ahead of the code.\\\\n\\\\nProse only in this section. A bullet list drops the connective tissue betwe", "expected": "document title returned", "input": "{\"tool\": \"get_document\", \"arguments\": {\"path\": \".archcore/use-redis-token-bucket.adr.md\"}}", "kind": "claim", "outcome": "Real input produced the expected result: Use Redis for the token bucket", "source": "Each command leaves plain Markdown in `.archcore/`, versioned with the code it describes.", "source_checked": true, "source_file": "README.md", "verdict": "yes" } ], "cycle_checks": [ { "claim": "Scheduled or looping main behavior", "command": "", "evidence": "Entry points inspected: README.md, archcore --help (init, mcp, status, hooks, update, doctor). Archcore is a CLI plus stdio MCP server; it has no built-in cron, scheduler, loop, --once/--now, or webhook trigger. Hooks are agent-driven (Claude Code/Codex/Copilot hook scripts in plugins/archcore/hooks and bin/), fired by the host agent per tool use, not by the product on a schedule. No product cycle to replay.", "expected": "identify the product trigger", "input": "source inspection", "kind": "cycle", "outcome": "Entry points inspected: README.md, archcore --help (init, mcp, status, hooks, update, doctor). Archcore is a CLI plus stdio MCP server; it has no built-in cron, scheduler, loop, --once/--now, or webhook trigger. Hooks are agent-driven (Claude Code/Codex/Copilot hook scripts in plugins/archcore/hooks and bin/), fired by the host agent per tool use, not by the product on a schedule. No product cycle to replay.", "source": "repository entry points", "verdict": "yes" } ], "details": { "argv": [ "", "mcp", "--project", "" ], "kind": "mcp", "scope": "product operation", "serverInfo": { "name": "archcore", "version": "0.13.1" }, "tool_result": { "content": [ { "text": "{\"path\":\".archcore/use-redis-token-bucket.adr.md\",\"category\":\"knowledge\",\"type\":\"adr\",\"title\":\"Use Redis for the token bucket\",\"status\":\"draft\",\"mtime\":\"2026-10-25T23:05:04.800758815Z\",\"content\":\"---\\ntitle: \\\"Use Redis for the token bucket\\\"\\nstatus: draft\\n---\\n\\n## Context\\n\\nState the trigger in 2 to 4 sentences: the situation now, the constraint or\\nmeasurement that turns it into a problem, and what breaks if nothing changes.\\nCite the evidence \u2014 @path/to/file, a metric, a commit, an external authority \u2014\\nor mark the claim [assumption] when the decision runs ahead of the code.\\n\\nProse only in this section. A bullet list drops the connective tissue between\\nthose sentences, and that tissue is what makes the decision follow from them.\\n\\n## Decision\\n\\nState the decision that was made.\\n\\n### Rationale\\n\\nExplain why this decision was chosen over alternatives:\\n\\n- Key factors that influenced the decision\\n- Trade-offs that were considered\\n- Assumptions that were made\\n\\n## Alternatives Considered\\n\\nOne numbered item per alternative, each naming what ruled it out. The reason is\\nthe part a later reader cannot reconstruct, and an alternative recorded without\\none reads as a preference rather than a choice.\\n\\n1. [Alternative] \u2014 rejected because [the measurement or constraint that ruled it out].\\n2. [Alternative] \u2014 rejected because [the measurement or constraint that ruled it out].\\n\\n## Consequences\\n\\n### Positive\\n\\n- Benefit 1: Description\\n- Benefit 2: Description\\n- Benefit 3: Description\\n\\n### Negative\\n\\n- Trade-off 1: Description and mitigation\\n- Trade-off 2: Description and mitigation\\n\\n### Risks\\n\\n- Risk 1: Description and mitigation strategy\\n- Risk 2: Description and mitigation strategy\\n\\n## Implementation Notes\\n\\nKey implementation considerations:\\n\\n- Migration path (if applicable)\\n- Dependencies affected\\n- Timeline considerations\\n\",\"source_id\":\"local\",\"source_kind\":\"local\"}", "type": "text" } ] }, "tool_result_checked": true, "tool_schema_bytes": 29556, "tools": [ "add_relation", "create_document", "get_document", "init_project", "install_host_config", "list_documents", "list_relations", "remove_document", "remove_relation", "search_documents", "update_document" ], "working": true }, "evidence_strength": "readme_checked", "example_checks": [ { "claim": "Connect your project with archcore init", "command": " init --project --yes", "deferred": false, "evidence": "\u2554\u2550\u2550\u2550\u2550\u2550\u2550\u2557 \n\u2551 \u2551 Archcore v0.13.1 \u2014 Git-native context for AI coding agents\n\u2551 \u2554\u2550\u2550\u2550\u2563 Spec-driven development & context engineering \n\u2551 \u2551 \u2554\u2550\u2563 https://archcore.ai \n\u255a\u2550\u2550\u2569\u2550\u2569\u2550\u255d \n\n \u2713 Created .archcore/ directory\n \u2713 Existing .archcore/settings.json kept\n No AI agent selected (non-interactive). Run 'archcore mcp install --agent ' later.\n\n Ready! Run 'archcore status' to verify.", "expected": "creates .archcore/ or keeps existing settings", "input": "archcore init --project --yes", "kind": "example", "outcome": "command completed successfully", "returncode": 0, "source": "README.md#install (step 2, archcore init)", "verdict": "yes" } ], "status": "PASS" }, "deviations": [ "README installer is curl|sh from archcore.ai; ran it after downloading to and reading it; telemetry disabled via DO_NOT_TRACK=1 and ARCHCORE_TELEMETRY_OPTOUT=1. Installed to (HOME-based default) rather than PATH; the probe uses the absolute path.", "Did not run archcore init inside the source workspace; used --project to keep the snapshot unmodified.", "The preflight path /inputs/coldpass-preflight.py does not exist; used as the instructions' first command specified.", "Probe and claim depend on a document created once during setup (create_document with filename use-redis-token-bucket); a second create returns 'file already exists', so the probe reads the existing document rather than creating it.", "", "No Node test suite run: plugins/archcore/tests/next-step.test.ts is a plugin hook test that needs a TS runner and is not a documented README command.", "No product cycle: archcore has no scheduler/cron/--once; hooks are driven by the host agent." ], "fault_owner": null, "main_blocker": null, "operation": "Install the Archcore CLI with the README installer, connect a project with archcore init, and read back a document through the archcore MCP stdio server (get_document).", "operation_scope": "Linux amd64 guest, Ubuntu, Python 3.12.3 (checker only). Archcore CLI v0.13.1 from archcore.ai install.sh. Project state lives in (a separate git-initialized dir, not the source snapshot). No AI agent host (Claude Code/Codex/Copilot) was present, so plugin install and agent slash commands were not exercised.", "outcome": "working", "plain_failure": null, "steps": [ "curl -fsSL https://archcore.ai/install.sh | bash (run as: DO_NOT_TRACK=1 ARCHCORE_TELEMETRY_OPTOUT=1 NODE_USE_ENV_PROXY=1 bash install.sh, telemetry opted out)", "archcore init (non-interactive, stdin null, no agent selected)", "archcore init --yes (rerun, kept existing settings)", "archcore mcp --project (stdio MCP server; create_document adr 'Use Redis for the token bucket' once during setup)", "get_document path .archcore/use-redis-token-bucket.adr.md (probe and claim replay)" ], "stuck_step": null, "suggested_fix": null }