spec-ctx 0.4.0 → 0.4.1

package/README.md

@@ -37,7 +37,7 @@ Six commands on a small project. Each clip is one command.
 
 OpenSpec names a requirement by its header text. A reword breaks every link to it. `adopt` gives each requirement and scenario an ID that stays. Run it once on a project that has specs. A new project gets IDs from `init --all`.
 
-![The adopt dry run lists each header and its new ID](docs/demo/clips/1-adopt.svg)
+![The adopt dry run lists each header and its new ID](https://raw.githubusercontent.com/sonqh/spec-ctx/main/docs/demo/clips/1-adopt.svg)
 
 You see: the dry run prints each header and the ID it gets. Then `adopt` writes them.
 
@@ -45,7 +45,7 @@ You see: the dry run prints each header and the ID it gets. Then `adopt` writes
 
 `status` shows the open change, the task counts, and the next command to run.
 
-![One change with two tasks done and one open](docs/demo/clips/2-status.svg)
+![One change with two tasks done and one open](https://raw.githubusercontent.com/sonqh/spec-ctx/main/docs/demo/clips/2-status.svg)
 
 You see: one change, two tasks done, one open, and the hint.
 
@@ -53,15 +53,15 @@ You see: one change, two tasks done, one open, and the hint.
 
 `task next` picks the first open task. It prints the requirement IDs the task cites and the Verify command, the shell command that proves the task is done.
 
-![Task add-mfa/1.3 with its requirement and Verify command](docs/demo/clips/3-next.svg)
+![Task add-mfa/1.3 with its requirement and Verify command](https://raw.githubusercontent.com/sonqh/spec-ctx/main/docs/demo/clips/3-next.svg)
 
 You see: task `add-mfa/1.3`, its requirement, and `grep -q totp src/login.ts`.
 
 ### 4. Only the spec this task needs
 
-`task start` prints the requirement, its scenarios, the task line, and the test that covers it. Nothing else. The last line is the size of that bundle next to the size of all spec text.
+`task start` prints the `context` answer for the task. It holds the task line and the requirements it cites with their scenarios. It holds the code and test lines that tag them, the decisions, and the deltas and designs of its change. All of it fits a character budget. The last line is the size of that text next to the size of all spec text.
 
-![The bundle of one task and its size footer](docs/demo/clips/4-start.svg)
+![The bundle of one task and its size footer](https://raw.githubusercontent.com/sonqh/spec-ctx/main/docs/demo/clips/4-start.svg)
 
 You see: 711 characters, 3.6 percent of the tree. The agent reads this instead of the tree.
 
@@ -69,15 +69,15 @@ You see: 711 characters, 3.6 percent of the tree. The agent reads this instead o
 
 `task done` runs the Verify command. On exit 1 the box stays open. On exit 0 the box flips to `[x]` and an evidence file records the files that were checked. The agent cannot tick the box by hand.
 
-![A refused done, the edit, and a passing done](docs/demo/clips/5-done.svg)
+![A refused done, the edit, and a passing done](https://raw.githubusercontent.com/sonqh/spec-ctx/main/docs/demo/clips/5-done.svg)
 
 You see: a red refusal, the edit, then a green done line.
 
 ### 6. Hand off to the next session
 
-`handoff write` puts the state in one file: the change, the done tasks, the current task, and the open questions. The next session starts from it. The questions live in `.ctx/questions.json`. `handoff ask` adds one and `handoff resolve <n>` answers it.
+`handoff write` puts the state in one file: the change, the done tasks, the current task, and the open questions. The next session starts from it. The questions live in `.ctx/questions.json`. `handoff ask` adds one and `handoff resolve <n>` answers it. A plan or archive question needs `--approve` or `--reject`, and a rejection keeps the stop.
 
-![The head of the handoff file](docs/demo/clips/6-handoff.svg)
+![The head of the handoff file](https://raw.githubusercontent.com/sonqh/spec-ctx/main/docs/demo/clips/6-handoff.svg)
 
 You see: the head of `.ctx/handoff.md`.
 
@@ -107,6 +107,26 @@ graph LR
 
 The ID is the stable key. A reword of a header does not change it. The index is the graph of IDs: tasks, requirements, scenarios, decisions, code tags, and test tags. `bundle` walks the edges to the configured depth inside a character budget. A decision record joins the graph through its `Refs:` line. The demo fixture has none, so the graph above shows no decision.
 
+### Ask for the context
+
+`specctx context <task|ID|path|change>` is the one command to run before you edit. It takes one input and prints the spec text that relates to it, with one provenance line for each block. A task gets the requirements it cites and the code and test lines that tag them. It also gets the decisions that name them, and the deltas and designs of its change. A file gets the requirements it tags and the code and test lines of those requirements, so its own tests appear. `task start` prints this same answer. This is the output for task `045/2.1` of this repository, with the text of each block left out:
+
+```text
+$ specctx context 045/2.1
+-- 045/2.1 | openspec/changes/archive/2026-10-30-045-query-by-file/tasks.md:11 | done | requested
+-- 045/D2 | openspec/changes/archive/2026-10-30-045-query-by-file/design.md:22 | archived | design named in 045/2.1
+-- 045/D3 | openspec/changes/archive/2026-10-30-045-query-by-file/design.md:23 | archived | design named in 045/2.1
+-- REQ-IDX-008 | openspec/specs/index/spec.md:133 | active | delta in change of 045/2.1
+-- REQ-IDX-009 | openspec/specs/index/spec.md:146 | active | delta in change of 045/2.1
+-- REQ-QRY-005 | openspec/specs/query/spec.md:76 | active | delta in change of 045/2.1
+-- REQ-QRY-006 | openspec/specs/query/spec.md:93 | active | ref from 045/2.1
+-- REQ-QRY-008 | openspec/specs/query/spec.md:106 | active | delta in change of 045/2.1
+-- src/graph/files.ts:1 | src/graph/files.ts:1 | active | implements REQ-QRY-006
+-- tests/query/file-edges.test.ts:1 | tests/query/file-edges.test.ts:1 | active | covers REQ-QRY-006
+```
+
+`bundle` gives the same answer for one input. It also takes several inputs, `--block`, and `--all`. When the text passes the budget, blocks drop in this order. First go the archived decisions and the entries two steps away. Then go the specs. Then go the code and test lines. The input and the requirements it cites stay.
+
 `specctx related` prints each neighbor of an ID and the reason for the link. This is the output for the demo project after its last step:
 
 ```text
@@ -118,6 +138,53 @@ code  src/login.ts:1  active  src/login.ts:1  implements REQ-UA-004
 test  tests/login.test.ts:1  active  tests/login.test.ts:1  covers REQ-UA-004
 ```
 
+### Ask about a file
+
+`related`, `context`, and `bundle` take a file path where they take an ID. Run it before you edit a file. The reason on each line tells you how strong the link is. `implements` comes from a code tag. `covers` comes from a `Covers:` tag. `mentions` comes from a path in backticks in a task line of an active change. This is the output for one file of this repository:
+
+```text
+$ specctx related src/util/hash.ts
+adr  ADR-004  active  decisions/ADR-004-tasks-md-is-the-view-evidence-is-the-record.md:1  ref from REQ-TASK-004
+req  REQ-CHK-003  active  openspec/specs/check/spec.md:64  ref to REQ-TASK-004
+req  REQ-STAT-002  active  openspec/specs/metrics/spec.md:19  ref to REQ-TASK-004
+req  REQ-TASK-004  active  openspec/specs/task-workflow/spec.md:89  implements, tag at src/util/hash.ts:1
+req  REQ-VER-004  active  openspec/specs/verify-runner/spec.md:37  ref to REQ-TASK-004
+scn  SCN-TASK-004-1  active  openspec/specs/task-workflow/spec.md:94  scenario of REQ-TASK-004
+task  002/3.3  done  openspec/changes/archive/2026-10-24-002-ids-and-tasks/tasks.md:17  task cites REQ-TASK-004
+```
+
+A path that is not a file in the repository exits 2. A file with no links prints `no links: <path>` and exits 0.
+
+### Ask about a change or a task
+
+`related`, `context`, and `bundle` take a change key or a folder name where they take an ID. A change reaches the requirements and scenarios of its delta specs and the capabilities they touch. This is the output for an archived change of this repository:
+
+```text
+$ specctx related 045
+req  REQ-IDX-008  active  openspec/specs/index/spec.md:133  delta in change/045
+req  REQ-IDX-009  active  openspec/specs/index/spec.md:146  delta in change/045
+req  REQ-QRY-005  active  openspec/specs/query/spec.md:76  delta in change/045
+req  REQ-QRY-006  active  openspec/specs/query/spec.md:93  delta in change/045
+req  REQ-QRY-008  active  openspec/specs/query/spec.md:106  delta in change/045
+scn  SCN-IDX-008-1  active  openspec/specs/index/spec.md:138  delta in change/045
+scn  SCN-IDX-008-2  active  openspec/specs/index/spec.md:142  delta in change/045
+scn  SCN-IDX-009-1  active  openspec/specs/index/spec.md:151  delta in change/045
+scn  SCN-IDX-009-2  active  openspec/specs/index/spec.md:155  delta in change/045
+scn  SCN-QRY-005-1  active  openspec/specs/query/spec.md:81  delta in change/045
+scn  SCN-QRY-005-2  active  openspec/specs/query/spec.md:85  delta in change/045
+scn  SCN-QRY-005-3  active  openspec/specs/query/spec.md:89  delta in change/045
+scn  SCN-QRY-006-1  active  openspec/specs/query/spec.md:98  delta in change/045
+scn  SCN-QRY-006-3  active  openspec/specs/query/spec.md:102  delta in change/045
+scn  SCN-QRY-008-1  active  openspec/specs/query/spec.md:111  delta in change/045
+scn  SCN-QRY-008-2  active  openspec/specs/query/spec.md:115  delta in change/045
+scn  SCN-QRY-008-3  active  openspec/specs/query/spec.md:119  delta in change/045
+scn  SCN-QRY-008-4  active  openspec/specs/query/spec.md:123  delta in change/045
+cap  cap/index  active  openspec/specs/index/spec.md:1  capability touched by change/045
+cap  cap/query  active  openspec/specs/query/spec.md:1  capability touched by change/045
+```
+
+A task reaches the same delta entries and the design decisions that its line names, for example `specctx related 045/2.1`. A requirement reaches its capability, and `related cap/query` lists the requirements of one capability. A modified requirement shows its main version beside the delta. The index stores only the `change` and `cap` entries and the design links of tasks. The other edges come from the folder structure when you ask. `query` hides the `change` and `cap` entries unless you pass `--kind change` or `--kind cap`.
+
 ## The life cycle for agents
 
 `init --all` installs one skill, `specctx-run`. The user types one prompt, such as "build this" or "continue". The skill loops until the change reaches the archive or a human must decide.
@@ -138,7 +205,7 @@ specctx gate --strict          # check, baseline, linters, openspec validate, le
 specctx handoff write          # .ctx/handoff.md for the next session
 ```
 
-The rule: the agent never opens `openspec/specs/` by hand. It uses `task start` or `bundle`. It never ticks a checkbox and never runs `/opsx:apply`. The skill is in [`skills/`](skills/). Install it alone with `npx skills add sonqh/spec-ctx` or `specctx init --skills`.
+The rule: the agent never opens `openspec/specs/` by hand. It uses `task start` or `context`. It never ticks a checkbox and never runs `/opsx:apply`. The skill is in [`skills/`](https://github.com/sonqh/spec-ctx/blob/main/skills/). Install it alone with `npx skills add sonqh/spec-ctx` or `specctx init --skills`.
 
 ## Install
 
@@ -158,7 +225,7 @@ The examples below use `specctx`. With `npx`, write `npx -y spec-ctx` in its pla
 
 ## Two paths
 
-[`docs/getting-started.md`](docs/getting-started.md) walks the fresh path on upstream OpenSpec with real output, for a developer who knows OpenSpec.
+[`docs/getting-started.md`](https://github.com/sonqh/spec-ctx/blob/main/docs/getting-started.md) walks the fresh path on upstream OpenSpec with real output, for a developer who knows OpenSpec.
 
 ### Fresh project
 
@@ -180,6 +247,7 @@ specctx baseline write
 specctx gate --strict
 ```
 
+- `adopt --dry-run` exits 1 when headers need IDs. A script with `set -e` must allow that code.
 - `adopt` gives an ID to every requirement and scenario header without one, in `openspec/specs/` and in the deltas of active changes. A delta header gets the main ID by its text or through its RENAMED line. A delta header that matches neither keeps its text and goes into `adopt-report.md`.
 - `adopt` derives a capability code from each folder name, for example `UA` for `user-auth`. It writes the codes into `spec-ctx.toml`. Rename a code there before the IDs spread.
 - `adopt` runs `check` before and after the write. A finding that the write added undoes every file. A second run prints `nothing to adopt`.
@@ -198,8 +266,9 @@ Both paths have an end to end test on upstream OpenSpec: `tests/paths/fresh.sh`
 | `status` | Active changes with task counts, the current task, and open questions. |
 | `next` | The next action of the loop from the files on disk: `plan`, `cite`, `start`, `work`, `fix`, `repair`, `gate`, `archive`, `ask`, or `stop`. `--json` prints one object. Flag: `--change`. |
 | `query <term...>` | Entries that hold every term. Flags: `--kind`, `--change`, `--limit`, `--all`. |
-| `related <ID>` | Linked entries with the reason for each link. Flags: `--depth`, `--all`. |
-| `bundle <ID...>` | Spec text of the IDs, their scenarios, decisions, and neighbors. Flags: `--budget`, `--depth`, `--all`. |
+| `context <task\|ID\|path\|change>` | The spec text, code lines, and test lines that relate to one input, in a budget. `task start` prints the same answer. Flags: `--budget`, `--depth`. |
+| `related <ID\|path\|change>` | Linked entries with the reason for each link. A file path or a change works too. Flags: `--depth`, `--all`. |
+| `bundle <ID\|path\|change...>` | The answer of `context` for several inputs. Flags: `--budget`, `--depth`, `--all`, `--block`. |
 | `check` | ID and link integrity. `--strict` adds test coverage and evidence. `retitle-needs-renamed` names a retitled delta header that needs a RENAMED pair before `openspec archive`. |
 | `id new`, `id retire` | Allocate and retire IDs. |
 | `task next`, `start`, `verify`, `done`, `reopen` | The task state machine. |
@@ -249,6 +318,9 @@ base = "origin/main"  # the ref whose merge base holds the baseline that gate co
 [loop]
 max_attempts = 3                        # Verify failures on one task before task done writes a question
 archive = "ask"                         # "ask" stops before archive for a human. "auto" lets the loop archive.
+
+[cli]
+run = "specctx"                         # the prefix of every command that spec-ctx prints. init writes npx -y spec-ctx@<version> under npx.
 ```
 
 With `brackets = true`, `init --schema` installs templates with `[REQ-<CAP>-NNN]` and `[SCN-<CAP>-NNN-N]` in the headers. The merged rules show the same form. Task citations, `Refs:` lines, and tags stay bare. After you change the value, run `specctx init --schema` again. It switches each template and rule that you did not edit. It keeps the ones you edited, unless you add `--force`.
@@ -317,8 +389,8 @@ A Verify command is code from the pull request. The boundary:
 
 ## Status
 
-- Published: `spec-ctx@0.4.0` on npm, with tags `v0.4.0` and `v0` and binaries on the GitHub release.
-- Archived: 001 to 025, 027 to 034, and 036 to 042. No change is open. See `CHANGELOG.md` for each release.
+- Published: `spec-ctx@0.4.1` on npm, with tags `v0.4.1` and `v0` and binaries on the GitHub release.
+- Archived: 001 to 025, 027 to 034, and 036 to 049. No change is open. See `CHANGELOG.md` for each release.
 - A version bump on `main` publishes the next release. See `docs/CONTRIBUTING.md`.
 - Upstream: the ID scheme is proposed as a comment on [OpenSpec#1967](https://github.com/Fission-AI/OpenSpec/issues/1967#issuecomment-6414314730). The schema bundle is submitted to the community table in [OpenSpec#2157](https://github.com/Fission-AI/OpenSpec/issues/2157). No pull request until a maintainer asks.
 
@@ -328,9 +400,9 @@ A Verify command is code from the pull request. The boundary:
 
 | change | context_ratio | evidence_ratio | stale_catches | cycle_time_median_s | debt |
 | --- | --- | --- | --- | --- | --- |
-| total | 0.0228 (n=116) | 0.93 (204/220) | 56 (of 106 reopens) | 61 (n=116) | 0 (0 baselined + 0 new) |
+| total | 0.0189 (n=135) | 0.94 (236/252) | 65 (of 118 reopens) | 60 (n=135) | 0 (0 baselined + 0 new) |
 
-- `context_ratio`: the size of one task bundle divided by the size of all spec text. Lower is better.
+- `context_ratio`: the size of one task bundle divided by the size of all spec text. Lower is better. The meaning changed on 2026-10-30 with change 047-context. Before, `task start` printed the cited requirements. Now it prints the `context` answer, which holds more. `bundle_chars` in the evidence of older tasks keeps the old size, so the two do not compare.
 - `evidence_ratio`: done tasks with a passed Verify command, divided by all done tasks. The rest have a manual note. Higher is better.
 - `stale_catches`: reopened tasks that `check --strict` flagged as stale, out of all reopened tasks. A higher share means the check found the drift before a person did.
 - `cycle_time_median_s`: the median number of seconds from `task start` to `task done`. Lower is faster.

package/dist/adopt/write.js

@@ -4,6 +4,7 @@ import { join } from "node:path";
 import { UsageExit } from "../cli/errors.js";
 import { CONFIG_FILE } from "../config.js";
 import { loadConfigOrExit } from "../cli/config.js";
+import { runCommand } from "../runcmd.js";
 import { dumps } from "../util/json.js";
 import { configText, withCaps } from "./config-write.js";
 export const REPORT_FILE = "adopt-report.md";
@@ -57,7 +58,7 @@ export function reportText(plan) {
     }
     return out.join("\n") + "\n";
 }
-export async function applyAdopt(root, _cfg, plan, io, json) {
+export async function applyAdopt(root, cfg, plan, io, json) {
     const cfgPath = join(root, CONFIG_FILE);
     const cfgBefore = existsSync(cfgPath) ? readFileSync(cfgPath, "utf8") : null;
     const cfgAfter = withCaps(configText(root), plan.caps);
@@ -120,8 +121,11 @@ export async function applyAdopt(root, _cfg, plan, io, json) {
         // REQ-CLI-007 only the codes in spec-ctx.toml changed: no header, no report
         if (!report)
             io.out(`wrote the codes to ${CONFIG_FILE}\nnothing to adopt\n`);
-        else
+        else {
             io.out(`adopted ${plan.edits.length} header(s) in ${files} file(s). Report: ${REPORT_FILE}\n`);
+            // REQ-ADOPT-002 adopt writes no baseline. The next step is a separate commit (044/D4, ADR-006 item 6).
+            io.out(`Commit the baseline with a subject that starts with baseline:.\nnext: ${runCommand(cfg.cli.run, "baseline write")}\n`);
+        }
     }
     return 0;
 }

package/dist/assets.generated.js

@@ -10,5 +10,5 @@ export const ASSETS = {
     "skills/specctx-bootstrap/SKILL.md": "LS0tCm5hbWU6IHNwZWNjdHgtYm9vdHN0cmFwCmRlc2NyaXB0aW9uOiBTdGFydCBhIHNlc3Npb24gaW4gYW4gT3BlblNwZWMgcHJvamVjdCB0aGF0IHVzZXMgc3BlYy1jdHguIFVzZSBhdCB0aGUgc3RhcnQgb2YgZXZlcnkgc2Vzc2lvbiwgYW5kIG9uICJzdGFydCIsICJib290c3RyYXAiLCBvciAid2hhdCBzaG91bGQgSSB3b3JrIG9uIi4gTG9hZHMgb25seSB0aGUgc3BlYyB0ZXh0IHRoYXQgdGhlIG5leHQgdGFzayBuZWVkcy4KLS0tCgpUaGlzIHNraWxsIGlzIG5vdyBgc3BlY2N0eC1ydW5gLiBSdW4gaXQuCg==",
     "skills/specctx-done/SKILL.md": "LS0tCm5hbWU6IHNwZWNjdHgtZG9uZQpkZXNjcmlwdGlvbjogQ2xvc2UgYSB0YXNrIGluIGFuIE9wZW5TcGVjIHByb2plY3QgdGhhdCB1c2VzIHNwZWMtY3R4LiBSdW5zIHRoZSBWZXJpZnkgY29tbWFuZCwgd3JpdGVzIGV2aWRlbmNlLCBzZXRzIHRoZSBjaGVja2JveCwgd3JpdGVzIHRoZSBoYW5kb2ZmLCBhbmQgc2hvd3MgdGhlIG1ldHJpY3MuIFVzZSB3aGVuIHRoZSB3b3JrIG9mIGEgdGFzayBpcyBjb21wbGV0ZSwgb3Igb24gImRvbmUiLCAiY2xvc2UgdGhlIHRhc2siLCBhbmQgImVuZCB0aGUgc2Vzc2lvbiIuCi0tLQoKVGhpcyBza2lsbCBpcyBub3cgYHNwZWNjdHgtcnVuYC4gUnVuIGl0Lgo=",
     "skills/specctx-next/SKILL.md": "LS0tCm5hbWU6IHNwZWNjdHgtbmV4dApkZXNjcmlwdGlvbjogUGljayBhbmQgc3RhcnQgdGhlIG5leHQgdGFzayBpbiB0aGUgbWlkZGxlIG9mIGEgc2Vzc2lvbiBpbiBhbiBPcGVuU3BlYyBwcm9qZWN0IHRoYXQgdXNlcyBzcGVjLWN0eC4gVXNlIGFmdGVyIHlvdSBjbG9zZSBhIHRhc2ssIG9yIG9uICJuZXh0IHRhc2siIGFuZCAid2hhdCBpcyBuZXh0Ii4KLS0tCgpUaGlzIHNraWxsIGlzIG5vdyBgc3BlY2N0eC1ydW5gLiBSdW4gaXQuCg==",
-    "skills/specctx-run/SKILL.md": "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",
+    "skills/specctx-run/SKILL.md": "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",
 };

package/dist/check/hints.js

@@ -1,14 +1,14 @@
 // REQ-CHK-004 the rules of check and one fix hint per rule. The check summary prints the hint of each rule it found.
 /** `hard` rules set exit 1. `advisory` rules print a warning and keep exit 0. */
 export const RULES = {
-    "missing-id": { level: "hard", hint: "Add an ID to the header. Run specctx adopt --dry-run to see the plan." },
+    "missing-id": { level: "hard", hint: "Add an ID to the header. Run {run} adopt --dry-run to see the plan." },
     "scenario-id-mismatch": { level: "hard", hint: "Give the scenario the number of its requirement." },
     "bad-ref": { level: "hard", hint: "Write only IDs on the Refs line, separated by commas." },
     "task-without-req": { level: "hard", hint: "Cite a requirement in brackets on the task line, for example [REQ-ABC-001]." },
     "task-without-verify": { level: "hard", hint: "Add Verify: and a command in backticks to the task line." },
-    "duplicate-id": { level: "hard", hint: "Give one of the two entries a new ID from specctx id new." },
+    "duplicate-id": { level: "hard", hint: "Give one of the two entries a new ID from {run} id new." },
     "duplicate-change-key": { level: "hard", hint: "Rename one change folder, so that each key names one change." },
-    "reused-id": { level: "hard", hint: "The ID is retired. Get a new ID from specctx id new." },
+    "reused-id": { level: "hard", hint: "The ID is retired. Get a new ID from {run} id new." },
     // REQ-CHK-007 the hint names the string literal cause and the fix (033/D3)
     "dangling-ref": {
         level: "hard",
@@ -20,13 +20,13 @@ export const RULES = {
     "unknown-cap": { level: "hard", hint: "Add the code to caps under [ids] in spec-ctx.toml." },
     uncovered: { level: "hard", hint: "Add a test with a Covers tag for the requirement." },
     "scenario-untested": { level: "hard", hint: "Name the scenario ID in a test with a // Covers: <SCN> tag." },
-    "done-without-evidence": { level: "hard", hint: "Run specctx task done. It runs the Verify command and writes the evidence." },
+    "done-without-evidence": { level: "hard", hint: "Run {run} task done. It runs the Verify command and writes the evidence." },
     // REQ-CHK-006 the hint names the why argument of task reopen (031/D8)
-    "stale-evidence": { level: "hard", hint: "Run specctx task reopen <task> <why>, then specctx task done <task>." },
-    "evidence-without-done": { level: "hard", hint: "Run specctx task done, or remove the evidence file." },
+    "stale-evidence": { level: "hard", hint: "Run {run} task reopen <task> <why>, then {run} task done <task>." },
+    "evidence-without-done": { level: "hard", hint: "Run {run} task done, or remove the evidence file." },
     // REQ-CHK-005 advisory in plain check. check --strict reports it as a hard finding (028/D5).
     "weak-verify": { level: "advisory", hint: "Name a test file or a command whose exit code proves the task." },
-    "manual-evidence": { level: "advisory", hint: "Give the task a Verify command and run specctx task done again." },
+    "manual-evidence": { level: "advisory", hint: "Give the task a Verify command and run {run} task done again." },
     // REQ-CHK-001 the hint is the block to paste. The finding line carries the ID and both titles (021/D3).
     // REQ-ADOPT-010 FROM copies the main header and TO copies the delta header, each in its own form (027/D4).
     "retitle-needs-renamed": {
@@ -34,6 +34,7 @@ export const RULES = {
         hint: "Add to the delta before the MODIFIED block. Write each ID as its header writes it, bare or in brackets:\n## RENAMED Requirements\n- FROM: `### Requirement: <ID> <old title>`\n- TO: `### Requirement: <ID> <new title>`",
     },
 };
-export function hintOf(rule) {
-    return RULES[rule]?.hint ?? "Read the check rules in the check spec.";
+/** REQ-CLI-010 `{run}` in a hint stands for the run prefix of `[cli] run` (043/D1). */
+export function hintOf(rule, run = "specctx") {
+    return (RULES[rule]?.hint ?? "Read the check rules in the check spec.").replaceAll("{run}", run);
 }

package/dist/cli/args.js

@@ -8,6 +8,7 @@ export const COMMAND_FLAGS = {
     query: { kind: "string", change: "string", all: "bool", limit: "int" },
     related: { depth: "int", all: "bool" },
     bundle: { budget: "int", depth: "int", all: "bool", block: "bool" },
+    context: { budget: "int", depth: "int" },
     check: { strict: "bool" },
     stats: { change: "string", write: "bool" },
     gate: { strict: "bool", markdown: "string", rerun: "bool", base: "string" },
@@ -20,7 +21,7 @@ export const COMMAND_FLAGS = {
     "task reopen": {},
     "handoff write": {},
     "handoff ask": { change: "string", plan: "bool" },
-    "handoff resolve": {},
+    "handoff resolve": { approve: "bool", reject: "bool" },
     adopt: { "dry-run": "bool" },
     "baseline write": {},
     "baseline prune": {},

package/dist/cli/commands.js

@@ -2,8 +2,10 @@
 import { resolve } from "node:path";
 import { UsageExit } from "./errors.js";
 import { findRoot, requireOpenSpec } from "./root.js";
+import { defaults } from "../config.js";
 import { hintOf } from "../check/hints.js";
 import { paint } from "./color.js";
+import { runCommand } from "../runcmd.js";
 /** check with warnings and the baseline filter. Strict mode adds the strict rules and the warnings. */
 async function runCheck(root, cfg, data, strict) {
     const { check } = await import("../check/check.js");
@@ -26,7 +28,7 @@ async function runCheck(root, cfg, data, strict) {
  * REQ-CHK-004 the findings, then a blank line, the summary line, and one line per rule with its hint.
  * A warning counts as advisory. The `baselined` line stays last (REQ-CHK-002).
  */
-export function checkText(r, tty) {
+export function checkText(r, tty, run = "specctx") {
     // REQ-CLI-008 every finding of check is hard, so every finding line is red. stale-evidence is one of them.
     // REQ-CHK-001 a finding with a detail, such as a retitle, carries it after the ID (021/D2)
     const body = r.findings.map((x) => paint("fail", `${x.file}:${x.line}  ${x.rule}  ${x.id}${x.detail ? `  ${x.detail}` : ""}\n`, tty)).join("")
@@ -62,7 +64,7 @@ export function checkText(r, tty) {
     const rules = [...count.keys()].sort();
     return body
         + "\n" + paint("dim", `${n} finding${n === 1 ? "" : "s"} in ${files.size} file${files.size === 1 ? "" : "s"} (${r.findings.length} hard, ${r.warnings.length} advisory)\n`, tty)
-        + rules.map((rule) => `${rule}  ${count.get(rule)}  ${own.get(rule)?.join("\n") ?? hintOf(rule)}\n`).join("")
+        + rules.map((rule) => `${rule}  ${count.get(rule)}  ${own.get(rule)?.join("\n") ?? hintOf(rule, run)}\n`).join("")
         + base;
 }
 export async function run(p, io) {
@@ -103,7 +105,7 @@ export async function run(p, io) {
             io.out(plan.unmatched.length ? planLines(plan).join("\n") + "\nnothing to adopt\n" : "nothing to adopt\n");
         // REQ-CLI-007 the dry run ends with the next step. REQ-ADOPT-001 the exit code stays 1 and the last line says why (017/D5).
         else
-            io.out(planLines(plan).join("\n") + `\nexit 1: ${plan.edits.length} headers need an ID. Next: specctx adopt\n`);
+            io.out(planLines(plan).join("\n") + `\nexit 1: ${plan.edits.length} headers need an ID. Next: ${runCommand(cfg.cli.run, "adopt")}\n`);
         return plan.edits.length ? 1 : 0;
     }
     if (cmd === "baseline write" || cmd === "baseline prune") {
@@ -163,7 +165,7 @@ export async function run(p, io) {
                 void check;
                 const data = buildIndex(root, cfg);
                 const r = await runCheck(root, cfg, data, strict);
-                return { code: r.findings.length ? 1 : 0, out: checkText(r, io.tty) };
+                return { code: r.findings.length ? 1 : 0, out: checkText(r, io.tty, cfg.cli.run) };
             },
         }, io);
     }
@@ -176,12 +178,17 @@ export async function run(p, io) {
     if (cmd === "query" && !f.change)
         need(1, "at least one term");
     if (cmd === "related") {
-        need(1, "one ID");
+        need(1, "one ID, path, or change");
         if (p.positional.length > 1)
-            throw new UsageExit("related takes one ID");
+            throw new UsageExit("related takes one ID, path, or change");
     }
     if (cmd === "bundle")
-        need(1, "at least one ID");
+        need(1, "at least one ID, path, or change");
+    if (cmd === "context") {
+        need(1, "one task, ID, path, or change");
+        if (p.positional.length > 1)
+            throw new UsageExit("context takes one task, ID, path, or change");
+    }
     if ((cmd === "status" || cmd === "check" || cmd === "next") && p.positional.length)
         throw new UsageExit(`${cmd} takes no arguments`);
     const { data, rebuilt } = loadIndex(root, cfg);
@@ -231,15 +238,35 @@ export async function run(p, io) {
         }
         case "related": {
             const { canon } = await import("../index/canonical.js");
-            const { related } = await import("../graph/related.js");
-            const id = p.positional[0];
+            const { related, mainVersion } = await import("../graph/related.js");
             const ids = canon(entries);
+            let id = p.positional[0];
+            let isPath = false;
             if (!ids.has(id) && !(all && entries.some((e) => e.id === id))) {
-                io.err(`unknown id ${id}\n`);
-                return 2;
+                // REQ-QRY-011 a change key or folder name is a change. REQ-QRY-005 any other input is a path in the repository.
+                const { resolveFileInput, changeInput } = await import("../graph/path.js");
+                const ch = changeInput(entries, id);
+                if (ch)
+                    id = ch;
+                else {
+                    const r = resolveFileInput(root, io.cwd, id);
+                    if ("error" in r) {
+                        io.err(`${r.error}\n`);
+                        return 2;
+                    }
+                    id = r.rel;
+                    isPath = true;
+                }
             }
             const rows = [];
             for (const [i, reason] of related(entries, id, f.depth ?? 1, all)) {
+                // REQ-QRY-010 the pair of a main version stands for the non-canonical copy in the main spec
+                if (!all && reason === `main version of ${i}`) {
+                    const m = mainVersion(entries, i);
+                    if (m)
+                        rows.push([m, reason]);
+                    continue;
+                }
                 const copies = all ? entries.filter((e) => e.id === i) : ids.has(i) ? [ids.get(i)] : [];
                 // canonical copy first, the rest in index order
                 copies.sort((a, b) => Number(!a.canonical) - Number(!b.canonical));
@@ -248,22 +275,27 @@ export async function run(p, io) {
             }
             if (json)
                 print(rows.map(([e, reason]) => ({ ...e, reason })));
+            else if (isPath && rows.length === 0)
+                io.out(`no links: ${id}\n`);
             else
                 io.out(rows.map(([e, r]) => `${e.kind}  ${e.id}  ${e.status}  ${e.file}:${e.line}  ${r}\n`).join(""));
             return 0;
         }
+        case "context":
         case "bundle": {
             const { canon } = await import("../index/canonical.js");
             const { bundle } = await import("../graph/bundle.js");
+            const { resolveInputs } = await import("../graph/input.js");
             const ids = canon(entries);
-            const unknown = p.positional.filter((i) => !ids.has(i));
-            if (unknown.length) {
-                io.err(`unknown id(s): ${unknown.join(", ")}\n`);
+            // REQ-QRY-005 an input that is no known ID is a path in the repository
+            const { wanted, paths, errors, taskId } = resolveInputs(root, io.cwd, entries, p.positional);
+            if (errors.length) {
+                io.err(errors.map((l) => l + "\n").join(""));
                 return 2;
             }
             // REQ-QRY-004 --block prints the requirement blocks verbatim and nothing else (034/D2)
             if (f.block) {
-                const notReq = p.positional.filter((i) => ids.get(i).kind !== "req");
+                const notReq = p.positional.filter((i) => ids.get(i)?.kind !== "req");
                 if (notReq.length) {
                     io.err(`not a requirement: ${notReq.join(", ")}\n`);
                     return 2;
@@ -276,9 +308,13 @@ export async function run(p, io) {
                     io.out(text);
                 return 0;
             }
-            const b = bundle(root, entries, p.positional, f.budget ?? cfg.bundle.budget, f.depth ?? cfg.bundle.depth, all);
+            // 047/D5 a task input passes the entry that findTask picks, so an active task wins over an archived one
+            const task = taskId === undefined ? undefined : (await import("../tasks/tasks.js")).findTask(root, entries, taskId).entry;
+            const b = bundle(root, entries, wanted, f.budget ?? cfg.bundle.budget, f.depth ?? cfg.bundle.depth, { all, task });
             if (json)
                 print(b);
+            else if (b.text === "" && wanted.every((w) => paths.has(w)))
+                io.out(wanted.map((w) => `no links: ${w}\n`).join(""));
             else
                 io.out(b.text);
             return 0;
@@ -310,7 +346,7 @@ export async function run(p, io) {
             if (json)
                 print({ baselined: r.baselined, count: findings.length, findings, warnings: r.warnings });
             else
-                io.out(checkText(r, io.tty));
+                io.out(checkText(r, io.tty, cfg.cli.run));
             return r.findings.length ? 1 : 0;
         }
     }
@@ -323,18 +359,28 @@ async function init(p, io) {
     const root = resolve(io.cwd, p.flags.root ?? ".");
     const all = !!p.flags.all;
     const r = runInit(root, { schema: all || !!p.flags.schema, skills: all || !!p.flags.skills, force: !!p.flags.force });
+    // REQ-CLI-010 init prints its next step with the run prefix of the config it wrote or kept (043/D1)
+    const { loadConfigOrExit } = await import("./config.js");
+    const run = (() => {
+        try {
+            return loadConfigOrExit(root).cli.run;
+        }
+        catch {
+            return defaults().cli.run;
+        }
+    })();
     if (p.flags.json)
         io.out(JSON.stringify(r, null, 2) + "\n");
     else if (r.nothing)
-        io.out("nothing to do\n" + r.notes.map((n) => `${n}\n`).join("") + paint("dim", "next: specctx status\n", io.tty));
+        io.out("nothing to do\n" + r.notes.map((n) => `${n}\n`).join("") + paint("dim", `next: ${runCommand(run, "status")}\n`, io.tty));
     else {
         io.out(r.written.map((w) => `wrote ${w}\n`).join(""));
         io.out(r.merged.map((m) => `merged ${m}\n`).join(""));
         io.out(r.kept.map((k) => `kept ${k}\n`).join(""));
         io.out(r.notes.map((n) => `${n}\n`).join(""));
         // REQ-CLI-007 init ends with the next step. REQ-DIST-002 a kept changed file points to --force (014/D2).
-        const again = ["specctx init", all ? "--all" : [p.flags.schema ? "--schema" : "", p.flags.skills ? "--skills" : ""].filter(Boolean).join(" "), "--force"].filter(Boolean).join(" ");
-        io.out(paint("dim", `next: ${r.changed.length ? again : "specctx status"}\n`, io.tty));
+        const again = [runCommand(run, "init"), all ? "--all" : [p.flags.schema ? "--schema" : "", p.flags.skills ? "--skills" : ""].filter(Boolean).join(" "), "--force"].filter(Boolean).join(" ");
+        io.out(paint("dim", `next: ${r.changed.length ? again : runCommand(run, "status")}\n`, io.tty));
     }
     return 0;
 }

package/dist/cli/help.js

@@ -38,19 +38,27 @@ export const TABLE = [
         exit: PLAIN, example: "specctx query lockfile --kind req",
     },
     {
-        name: "related", group: "Read", args: "<ID>", summary: "Show the entries that link to an ID.",
+        name: "context", group: "Read", args: "<task|ID|path|change>", summary: "Print what an agent needs before it edits a task, an ID, a file, or a change.",
+        flags: [
+            ["--budget <n>", "Stop at n characters. The default is 8000."],
+            ["--depth <n>", "Follow links n steps. The default is 1."],
+        ],
+        exit: [OK, ["2", "A usage error, an unknown ID, or a bad path. It takes one input. Use `bundle` for several inputs or for `--block`."], INTERNAL], example: "specctx context 045/2.1",
+    },
+    {
+        name: "related", group: "Read", args: "<ID|path|change>", summary: "Show the entries that link to an ID, a file, or a change.",
         flags: [["--depth <n>", "Follow links n steps. The default is 1."], ["--all", "Show all copies of an ID."]],
-        exit: [OK, ["2", "A usage error or an unknown ID."], INTERNAL], example: "specctx related REQ-IDX-001",
+        exit: [OK, ["2", "A usage error, an unknown ID, or a path that is not a file in the repository."], INTERNAL], example: "specctx related src/graph/bundle.ts",
     },
     {
-        name: "bundle", group: "Read", args: "<ID...>", summary: "Print the spec text for the IDs and their links.",
+        name: "bundle", group: "Read", args: "<ID|path|change...>", summary: "Print the spec text for the IDs, files, or changes and their links.",
         flags: [
             ["--budget <n>", "Stop at n characters. The default is 8000."],
             ["--depth <n>", "Follow links n steps. The default is 1."],
             ["--all", "Show all copies of an ID."],
             ["--block", "Print only the requirement block, as it sits in the spec file. Each ID must be a requirement."],
         ],
-        exit: [OK, ["2", "A usage error, an unknown ID, or an ID that is not a requirement."], INTERNAL], example: "specctx bundle REQ-IDX-001 --budget 3000",
+        exit: [OK, ["2", "A usage error, an unknown ID, a bad path, or an ID that is not a requirement."], INTERNAL], example: "specctx bundle REQ-IDX-001 --budget 3000",
     },
     {
         name: "check", group: "Read", args: "", summary: "Find ID and link errors.",
@@ -125,7 +133,11 @@ export const TABLE = [
     },
     {
         name: "handoff resolve", group: "Write", args: "<n> <answer>", summary: "Move open question n to the resolved list.",
-        flags: [], exit: PLAIN, example: "specctx handoff resolve 1 \"yes, for one release\"",
+        flags: [
+            ["--approve", "Approve a plan or archive question. A plan or archive question needs this or --reject."],
+            ["--reject", "Reject a plan or archive question. The stop stays."],
+        ],
+        exit: PLAIN, example: "specctx handoff resolve 1 --approve \"yes, for one release\"",
     },
     // Setup
     {