ChatWalaʻau v0.113.0

This release adds one new feature to ChatWalaʻau: the prompt dump now shows the agent's tool surface -- every tool it has, every tool it does not, and why.

No breaking changes. No operator action required. Everything here is additive and only visible when the optional PROMPT_DUMP_ENABLED debug toggle is on, which is off by default. No new environment variable, no API change, no UI change, no stored-data change. With the toggle off, this release behaves identically to v0.112.2.

Highlights

Prompt dump: see the tools the agent actually has (v0.113.0)

• With PROMPT_DUMP_ENABLED=true, each dump file gains a Tool Surface section listing every built-in function, every MCP server and tool, and every Skill. Until now the dump showed the system prompt and the conversation but not the tools -- even though the tool definitions are sent to the model in the same request.
• For anything missing, the dump says why, and each reason points at the screen that owns the gate: an environment setting (the row names the variable, e.g. CODING_ENABLED=false), the Declarative Agent's tool allow-list, the MCP Tool Manager, the Skills Manager, or an MCP server that is not connected.
• It is a self-check, not just a listing. The tools are read from the agent that is actually about to run and compared against what your configuration says they should be. When the two disagree with no explanation, the row is flagged MISMATCH -- so a tool you enabled that quietly did not make it into the built agent becomes visible instead of silent.
• A not-connected MCP server is called out as a WARNING. While a server is unconnected its tool names are unknown, so a per-tool allow-list cannot be applied and every tool of that server will be exposed once it connects.
• The section is written on every turn (the system prompt is still written on the first turn only), because tool state can change mid-session from the MCP manager, the Skills manager, or by switching Declarative Agent. A short digest line shows at a glance which turn the surface changed on.
• Producing the report never modifies anything: it does not change MCP tool exposure, does not touch connections, and does not rebuild agents. If it fails, the dump is still written without the section and the chat is unaffected.

Example:

### Built-in functions (4 active / 13 excluded)

- file_write                      ACTIVE    approval: always_require
- bash_execute                    EXCLUDED  allowlist
- rag_search                      EXCLUDED  settings (CHROMA_DIR unset or no embeddings offering)
- manage_cron                     EXCLUDED  settings (CRON_ENABLED=false)

### MCP (1 active server / 1 excluded / 1 warning)

- github                          ACTIVE    2/14 tools
    - create_issue                ACTIVE
    - delete_repo                 EXCLUDED  mcp override
- filesystem                      EXCLUDED  allowlist (server not selected)
- slack                           WARNING   not connected; allow-list not applied

Known issue

• Running a declarative Workflow can narrow the MCP tools of the ordinary chat agent, because workflow node agents share MCP tool objects with the chat agent. This release makes the problem observable (it appears as MISMATCH rows after a workflow run) but does not fix it. A fix is tracked for a following release. If you do not run declarative Workflows, you are not affected.

Included in this release

This release includes implementation work from:

• v0.113.0 through v0.113.0

Notes

• Internal design management content and architecture control artifacts are intentionally omitted from this release note.

Installation

See the repository README for setup and usage instructions.

Version

• Release version: 0.113.0
• Previous release: 0.112.2