From 2dacf3de0f4b17eb29f04c057bf9674918172bcb Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Mon, 7 Sep 2026 10:32:19 +0000 Subject: [PATCH] docs: update glossary with weekly full scan (logs storage budget, dynamic enclave delegation) - Add Logs Storage Budget (--max-storage) documenting gh aw logs selective pruning/storage cap (#59049) - Add Dynamic Enclave Delegation Controller documenting enclaves[].agent.dynamic mode (#59046) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --- docs/src/content/docs/reference/glossary.md | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/docs/src/content/docs/reference/glossary.md b/docs/src/content/docs/reference/glossary.md index ed890692841..3beb349457e 100644 --- a/docs/src/content/docs/reference/glossary.md +++ b/docs/src/content/docs/reference/glossary.md @@ -91,6 +91,10 @@ An MCP server field that controls startup criticality. By default every configur A response pattern used by the `gh aw` MCP server's `logs` tool when a gateway timeout or token budget guardrail prevents returning a complete result set. Instead of failing the call, the tool returns the JSON data it collected so far with `partial: true` and continuation parameters (such as `after_run_id`/`before_run_id`) so the caller can request the remaining results in a follow-up call. See [`gh aw` as an MCP Server](/gh-aw/reference/gh-aw-as-mcp-server/). +### Logs Storage Budget (`--max-storage`) + +A `gh aw logs` flag that caps total on-disk storage in MB for downloaded run artifacts and cache data. When downloads approach the configured limit, the command performs selective pruning: it removes non-essential cache data (such as prunable log-cache files not needed by already-completed runs) before falling back to stopping further downloads with `errLogsStorageLimitReached`. Reserving space per download and recording pruned bytes lets long-running multi-workflow log collection stay within a fixed disk budget instead of exhausting it. See [Monitoring Costs with `gh aw logs`](/gh-aw/reference/cost-management/#monitoring-costs-with-gh-aw-logs). + ### QMD Documentation Search (`qmd:`) A built-in tool that provides vector similarity search over documentation files. Configured via `tools.qmd:` in frontmatter, the `qmd` tool runs [tobi/qmd](https://github.com/tobi/qmd) as an MCP server so agents can find relevant documentation by natural language query. The search index is built in a dedicated indexing job (which has `contents: read`) and shared with the agent job via `actions/cache`, so the agent job does not need `contents: read`. Supports indexing from repository checkouts, GitHub code search queries, and cache-only read-only mode. See [QMD Documentation Search](/gh-aw/experimental/qmd/). @@ -163,6 +167,10 @@ A built-in tool that connects agentic workflows to Atlassian's official remote R A top-level frontmatter array that enables finite-disclosure access to approved repositories from within a public-facing workflow. Repository sensitivities are `public`, `trusted`, `internal`, `confidential`, or `sealed`; `trusted` is unmetered and permits free-form strings only inside a declared strict structured response schema, while the other sensitivities remain finite-schema-only. The compiler registers `enclave_run_script` or `enclave_run_agent` tools from the keyed `script`/`agent` entries present on the `awf-enclave` MCP route, compiled through [mcpg](#mcp-gateway) with run-scoped capability handoff, timeout derivation, and network validation. Enclaves require AWF network isolation, which every supported `sandbox.agent.runtime` profile provides, so the compiler can launch mcpg in bridge mode. Agent enclaves may opt into `github.cli: issues-read-v1`, which uses a distinct identity on the shared compiler-owned mcpg gateway and permits only `list_issues` and `issue_read` for the configured trusted repositories. AWF privately stages this identity and connects the enclave directly to `/mcp/github`; the enclave receives neither a GitHub token nor a `gh` executable. See [Private Repository Enclaves](/gh-aw/experimental/enclaves/). +### Dynamic Enclave Delegation Controller (`enclaves[].agent.dynamic`) + +An [Enclaves](#enclaves-enclaves) `dynamic:` mode that lets an agent enclave select one admitted repository at invocation time (via `allowed-owners`/`allowed-repositories`) instead of enumerating every repository statically in `repos:`. The compiler emits a fixed-sensitivity policy envelope with finite resource limits, total quotas, audit labels, and an absolute `expires-at` timestamp, then starts mcpg's `github-repository-delegation-v1` controller and hands AWF a host-private delegation control endpoint that is excluded from both primary and enclave agent environments. The effective envelope expiry is `min(expires-at, job-start + enclave timeout)`. Requires AWF `v0.28.14`+ and mcpg `v0.4.17`+ (the first release accepting the controller's atomic bootstrap configuration). Dynamic mode is agent-only; script enclaves remain static and must declare `repos`. See [Private Repository Enclaves](/gh-aw/experimental/enclaves/). + ### MCP Scripts Custom MCP tools defined inline in workflow frontmatter using JavaScript or shell scripts. Enables lightweight tool creation while maintaining controlled secret access. Tools are generated at runtime and mounted as an MCP server with typed input parameters, default values, and environment variables. Configured via `mcp-scripts:` section. Use the `dependencies:` field to declare third-party npm or PyPI packages installed before first invocation. See [MCP Scripts Reference](/gh-aw/reference/mcp-scripts/).