docs/MCP.md describes the feed and returns tools: `feed_chips`, `next_steps`, `returns`, `thought` #5303

docs/MCP.md now has a row in the tool table for each of feed_chips, next_steps, returns and thought. I didn't run the tests, since only a docs file changed, and I didn't check that a check would pass. A bun test I started was still running in the background when I pushed, so I have no result from it.

• feed_chips: sits right after feed. Its row says what it hides, and that it switches a chip when given kind (kind or person), value and hidden. It is the only one of the four ticked as writing.
• thought: sits after tell, since a tell without a repo is where the id comes from. It reads a thought's placement and its repeat/supersede links, and answers not_found for a thought that isn't yours or doesn't exist.
• next_steps and returns: sit after now. next_steps lists a spawned forge's first three steps. returns reports what the crew's tokens bought over 1–90 days, optionally for one repo.

Each description comes from the tool's text and handler in src/http/mcp.ts. There is no source change, and test/mcp-docs.test.ts doesn't exist yet, so there was no allowlist to edit.

Closes #5281

docs/MCP.md now has a row in the tool table for each of feed_chips, next_steps, returns and thought. I didn't run the tests, since only a docs file changed, and I didn't check that a check would pass. A bun test I started was still running in the background when I pushed, so I have no result from it.

• feed_chips: sits right after feed. Its row says what it hides, and that it switches a chip when given kind (kind or person), value and hidden. It is the only one of the four ticked as writing.
• thought: sits after tell, since a tell without a repo is where the id comes from. It reads a thought's placement and its repeat/supersede links, and answers not_found for a thought that isn't yours or doesn't exist.
• next_steps and returns: sit after now. next_steps lists a spawned forge's first three steps. returns reports what the crew's tokens bought over 1–90 days, optionally for one repo.

Each description comes from the tool's text and handler in src/http/mcp.ts. There is no source change, and test/mcp-docs.test.ts doesn't exist yet, so there was no allowlist to edit.

The full gate on land was red; the pull is open again. Red tests:

• the allowlist names only undocumented tools
• the allowlist names only undocumented tools [0.61ms] [shard 762]

Red on top of #5192, which landed first from the same merge-queue candidate: red on its own on that base, or only together with them. Bring this pull up to date with the base and push.

The full gate on land was red; the pull is open again. Red tests:

• the allowlist names only undocumented tools
• the allowlist names only undocumented tools [0.60ms] [shard 762]

The red test is fixed and pushed. I ran test/mcp-docs.test.ts on its own after the change and it passes (3 pass, 0 fail). I didn't run the rest of the suite.

The failing test was "the allowlist names only undocumented tools". test/mcp-docs.test.ts landed on main with next_steps, thought, feed_chips and returns still on its allowlist of undocumented tools, and this pull had since documented them. I merged origin/main into the branch and removed those four names from the allowlist.

I did not pad the new docs/MCP.md rows to the table column widths, as juniper suggested. Several existing rows are already unpadded, so the table isn't aligned today. The other suggestion, a test that fails when a tool has no docs row, is the test/mcp-docs.test.ts that main now has.