This commit was merged in pull request #669.
This commit is contained in:
@@ -2,6 +2,7 @@
|
||||
|
||||
## Unreleased
|
||||
|
||||
- Load no MCP server in the six headless `claude -p` calls that deny the model tools, so text an outsider wrote cannot reach one.
|
||||
- `do-epic` starts an epic's runnable issues in order of score alone, not in Jira's backlog order. It scores each open issue that holds no scores, the ones people filed included, and comments on the issue why.
|
||||
- do-epic and `todo-item add` file issues as the Jira issue types the settings `jira-finding-type` and `jira-review-type` name, Task by default, and refuse before filing where the Jira project lacks the type, naming the types it has.
|
||||
- do-epic names the team's own label for each repo, from the `[label-repos]` table, when it tells a chunk's worker how to file a finding or a release review, and keeps `repo:<name>` for a repo the table doesn't name.
|
||||
|
||||
@@ -15,8 +15,7 @@
|
||||
| 454 | 0.15.0 | | | **Start a do-epic chunk's issue as the team's board expects, by settings: assigned to the account do-epic works as, moved by a start transition and added to a board's active sprint.** | 3 | 2 | 2 | 5 | 4 | 1.6 by 4 | 1.1 |
|
||||
| 472 | 0.15.0 | 454 | owner | **Let a project's config name a working branch per epic, such as `TW-12345`, that chunks branch off, merge into and keep level with `main`, so a person merges the finished epic into `main` in one pull request.** | 4 | 2 | 3 | 7 | 7 | 1.5 by 6, 1.6 by 7 | 1.1 |
|
||||
| 467 | 0.15.0 | 454, 472 | question | **Cut 0.15.0 as soon as items 449, 450, 453, 454, 468, 470 and 472 have merged, before its release review, and run that review in 0.15.1.** | 1 | 1 | 1 | 1 | 1 | | 0.0 |
|
||||
| 473 | 0.15.0 | | security | **Load no MCP server in a headless claude call that is denied tools, so text an outsider wrote cannot reach one.** | 2 | 2 | 2 | 4 | 3 | 4 by 4 | 5.0 |
|
||||
| 476 | 0.15.0 | 473 | security | **Load no MCP server in the bench's headless claude calls that are denied tools, as item 473 did for the loop's six.** | 2 | 2 | 1 | 3 | 2 | 4 by 3 | 4.2 |
|
||||
| 476 | 0.15.0 | | security | **Load no MCP server in the bench's headless claude calls that are denied tools, as item 473 did for the loop's six.** | 2 | 2 | 1 | 3 | 2 | 4 by 3 | 4.2 |
|
||||
| 469 | 0.15.1 | | | **Move a do-epic parent issue whose sub-tasks a person filed to Release Ready once every sub-task has landed, and to Done once every sub-task is done, by transitions its config names.** | 3 | 2 | 2 | 5 | 4 | 1.2 by 4, 1.5 by 4, 1.6 by 3 | 3.1 |
|
||||
| 446 | 0.16.0 | | question | **Kill the process a stuck check names where its verdict is `stuck`, once the human says the recorded verdicts match what they would have done.** | 2 | 2 | 3 | 5 | 3 | 1.5 by 3 | 1.0 |
|
||||
| 462 | 0.16.0 | | | **Run under `mutate-changes` the tests that today fail in mutmut's copy of the tree, so a mutant of a line only they cover is killed.** | 2 | 1 | 2 | 3 | 3 | code health by 3 | 4.7 |
|
||||
@@ -137,36 +136,6 @@ Touches:
|
||||
|
||||
- `CHANGELOG.md`: `## Unreleased` (small)
|
||||
|
||||
### 473. Load no MCP server in a headless claude call that is denied tools, so text an outsider wrote cannot reach one.
|
||||
|
||||
Exposure: six headless `claude -p` calls deny the model tools with `--tools ""`, or with a read-only list, while their prompts hold text an outsider can write. That text is a Jira issue's summary and description, or a worker's transcript. `claude --help` says `--tools` sets only the built-in tools. MCP servers from the user's or the project's config still load, so a prompt injection in that text can call an MCP tool. What it reaches is whatever those servers can do, such as reading or posting in another client's tracker or chat. `--strict-mcp-config` loads only the servers `--mcp-config` names, and none where it names none.
|
||||
|
||||
Call sites:
|
||||
|
||||
- `bin/do_epic/repo_pick.py:105`, in `picked`.
|
||||
- `bin/do_epic/issue_scores.py:36`, in `asked`.
|
||||
- `bin/do_epic/split.py:126`, in `prediction`, which allows Read, Grep and Glob.
|
||||
- `bin/agent_cycle/judge.py:211`, in `judge_args`.
|
||||
- `bin/agent_cycle/stuck_workers.py:107`, in `investigated`.
|
||||
- `bin/review_replay/replay.py:194`, in `match`.
|
||||
|
||||
Fix: pass `--strict-mcp-config`, without `--mcp-config`, beside `--tools` at each call site. Add a test per call site that the flag is passed. Found by item 470's security pass; it is older than that chunk.
|
||||
|
||||
Touches:
|
||||
|
||||
- `bin/do_epic/repo_pick.py`: `picked` (small)
|
||||
- `bin/do_epic/issue_scores.py`: `asked` (small)
|
||||
- `bin/do_epic/split.py`: `prediction` (small)
|
||||
- `bin/agent_cycle/judge.py`: `judge_args` (small)
|
||||
- `bin/agent_cycle/stuck_workers.py`: `investigated` (small)
|
||||
- `bin/review_replay/replay.py`: `match` (small)
|
||||
- `bin/do_epic/test_repo_pick.py`: `RepoPick`
|
||||
- `bin/do_epic/test_issue_scores.py`
|
||||
- `bin/do_epic/test_split.py`: `Split_`
|
||||
- `bin/agent_cycle/test_judge.py`: `JudgePrompt`
|
||||
- `bin/agent_cycle/test_stuck_workers.py`: `StuckWorkers`
|
||||
- `bin/review_replay/test_review_replay.py`
|
||||
|
||||
### 476. Load no MCP server in the bench's headless claude calls that are denied tools, as item 473 did for the loop's six.
|
||||
|
||||
Exposure: item 473 added `--strict-mcp-config`, without `--mcp-config`, beside `--tools` at six headless `claude -p` calls in `bin/` that deny the model tools while their prompt holds text an outsider can write. Without the flag, MCP servers from the user's or the project's config still load, so a prompt injection in that text can call an MCP tool. Two more calls in the repo's bench folders deny tools the same way and still lack the flag:
|
||||
|
||||
Reference in New Issue
Block a user