pantheon-opencode 2.1.2 → 2.1.4

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (42) hide show
  1. package/AGENTS.global.md +7 -7
  2. package/AGENTS.md +1 -1
  3. package/CHANGELOG.md +77 -0
  4. package/README.md +10 -0
  5. package/package.json +7 -4
  6. package/plugin.json +1 -1
  7. package/pyproject.toml +1 -1
  8. package/scripts/build-plugin.mjs +35 -9
  9. package/scripts/build-roster.mjs +139 -0
  10. package/scripts/doctor.mjs +111 -2
  11. package/scripts/install/opencode.mjs +16 -0
  12. package/scripts/sync-tui.mjs +2 -0
  13. package/scripts/test-slice.mjs +352 -0
  14. package/skills-lock.json +1 -1
  15. package/src/agents/aphrodite.md +8 -6
  16. package/src/agents/apollo.md +9 -7
  17. package/src/agents/athena.md +21 -18
  18. package/src/agents/demeter.md +7 -5
  19. package/src/agents/gaia.md +7 -8
  20. package/src/agents/hephaestus.md +7 -5
  21. package/src/agents/hermes.md +20 -16
  22. package/src/agents/iris.md +10 -8
  23. package/src/agents/mnemosyne.md +17 -19
  24. package/src/agents/nyx.md +7 -5
  25. package/src/agents/prometheus.md +7 -5
  26. package/src/agents/talos.md +11 -5
  27. package/src/agents/themis.md +11 -4
  28. package/src/agents/zeus.md +23 -2
  29. package/src/instructions/agent-return-format.instructions.md +1 -1
  30. package/src/instructions/test-economy.instructions.md +6 -6
  31. package/src/mcp/pantheon_vision.py +24 -0
  32. package/src/pantheon/auto-permissions.ts +328 -0
  33. package/src/pantheon/deepwork/deepwork-tool.ts +14 -0
  34. package/src/pantheon/delegation-enforce.ts +43 -12
  35. package/src/pantheon/native-vision.ts +54 -2
  36. package/src/pantheon/tools.ts +102 -3
  37. package/src/plugin/index.js +342 -16
  38. package/src/plugin.ts +7 -0
  39. package/src/plugins/tui/dist/tui.js +5 -5
  40. package/src/plugins/tui/package-lock.json +2 -2
  41. package/src/plugins/tui/package.json +1 -1
  42. package/src/skills/test-slicing/SKILL.md +44 -15
@@ -0,0 +1,352 @@
1
+ #!/usr/bin/env node
2
+ /**
3
+ * Test Slice Engine — Observable & Diff-Aware Slice Selector.
4
+ *
5
+ * Implements Issue #437:
6
+ * 1. Prefer repository-level focused selectors (e.g. `test:quick`, `test:fast`) over path guessing.
7
+ * 2. Map modified source files directly to narrow test slices (<10s).
8
+ * 3. Report mapping gaps with a bounded static/lint fallback instead of escalating to full suites.
9
+ * 4. Support failure/fix loops with `--failed-only <pattern>`.
10
+ * 5. Emit formatted task receipts: command, target count, duration, selection rationale.
11
+ *
12
+ * @module scripts/test-slice
13
+ */
14
+
15
+ import { execSync, spawnSync } from 'node:child_process'
16
+ import { existsSync, readFileSync } from 'node:fs'
17
+ import { basename, dirname, extname, join } from 'node:path'
18
+
19
+ /**
20
+ * Known repo-level fast slice script names in package.json, ordered by preference.
21
+ */
22
+ const REPO_SLICE_SCRIPTS = [
23
+ 'test:quick',
24
+ 'test:fast',
25
+ 'test:slice',
26
+ 'test:local',
27
+ 'test:unit',
28
+ 'test-local',
29
+ 'precheck',
30
+ ]
31
+
32
+ /**
33
+ * Extensions representing source files that need test coverage.
34
+ */
35
+ const SOURCE_EXTENSIONS = new Set([
36
+ '.ts',
37
+ '.tsx',
38
+ '.js',
39
+ '.jsx',
40
+ '.mjs',
41
+ '.cjs',
42
+ '.py',
43
+ '.go',
44
+ '.rs',
45
+ ])
46
+
47
+ /**
48
+ * Extensions representing test files.
49
+ */
50
+ const TEST_EXTENSIONS = new Set([
51
+ '.test.ts',
52
+ '.test.tsx',
53
+ '.test.js',
54
+ '.test.jsx',
55
+ '.test.mjs',
56
+ '.spec.ts',
57
+ '.spec.js',
58
+ ])
59
+
60
+ /**
61
+ * Inspect package.json in directory for an existing focused slice script.
62
+ */
63
+ export function findRepoSliceScript(cwd = process.cwd()) {
64
+ const pkgPath = join(cwd, 'package.json')
65
+ if (!existsSync(pkgPath)) return null
66
+
67
+ try {
68
+ const pkg = JSON.parse(readFileSync(pkgPath, 'utf8'))
69
+ const scripts = pkg.scripts || {}
70
+ for (const name of REPO_SLICE_SCRIPTS) {
71
+ if (scripts[name]) {
72
+ return { name, command: `npm run ${name}` }
73
+ }
74
+ }
75
+ } catch {
76
+ // Ignore invalid json
77
+ }
78
+ return null
79
+ }
80
+
81
+ /**
82
+ * Get changed files in git working tree.
83
+ */
84
+ export function getChangedFiles(cwd = process.cwd()) {
85
+ try {
86
+ const stdout = execSync('git status --porcelain', { cwd, encoding: 'utf8' })
87
+ const files = []
88
+ for (const line of stdout.split('\n')) {
89
+ const trimmed = line.trim()
90
+ if (!trimmed) continue
91
+ // Format: XY path or XY orig -> path
92
+ const parts = trimmed.substring(2).trim().split(' -> ')
93
+ const filePath = parts[parts.length - 1]
94
+ if (filePath && !files.includes(filePath)) {
95
+ files.push(filePath)
96
+ }
97
+ }
98
+ return files
99
+ } catch {
100
+ return []
101
+ }
102
+ }
103
+
104
+ /**
105
+ * Map a single source or test file to its narrow test slice command.
106
+ */
107
+ export function mapFileToSlice(filePath, cwd = process.cwd()) {
108
+ const normalized = filePath.replace(/\\/g, '/')
109
+ const ext = extname(normalized)
110
+ const base = basename(normalized)
111
+
112
+ // 1. If already a test file, run it directly
113
+ if (
114
+ TEST_EXTENSIONS.has(ext) ||
115
+ normalized.includes('.test.') ||
116
+ normalized.includes('.spec.') ||
117
+ base.startsWith('test_')
118
+ ) {
119
+ if (ext === '.py') {
120
+ return {
121
+ command: `pytest ${normalized} -q --tb=short`,
122
+ testFile: normalized,
123
+ type: 'python-test',
124
+ selection: `direct test file: ${normalized}`,
125
+ }
126
+ }
127
+ if (ext === '.ts' || ext === '.tsx') {
128
+ return {
129
+ command: `npx tsx ${normalized}`,
130
+ testFile: normalized,
131
+ type: 'ts-test',
132
+ selection: `direct test file: ${normalized}`,
133
+ }
134
+ }
135
+ return {
136
+ command: `node --test ${normalized}`,
137
+ testFile: normalized,
138
+ type: 'node-test',
139
+ selection: `direct test file: ${normalized}`,
140
+ }
141
+ }
142
+
143
+ // 2. Map source file to candidate test locations
144
+ const stem = base.replace(/\.(ts|tsx|js|jsx|mjs|cjs|py)$/, '')
145
+ const dir = dirname(normalized)
146
+
147
+ const candidatePaths = [
148
+ // Same directory
149
+ join(dir, `${stem}.test${ext}`).replace(/\\/g, '/'),
150
+ join(dir, `${stem}.spec${ext}`).replace(/\\/g, '/'),
151
+ // tests/ directory mirroring
152
+ join('tests', `${stem}.test.mjs`).replace(/\\/g, '/'),
153
+ join('tests', `${stem}.test.ts`).replace(/\\/g, '/'),
154
+ join('tests', 'pantheon', `${stem}.test.ts`).replace(/\\/g, '/'),
155
+ join('tests', `test_${stem}.py`).replace(/\\/g, '/'),
156
+ join('tests', `${stem}.test.js`).replace(/\\/g, '/'),
157
+ ]
158
+
159
+ for (const candidate of candidatePaths) {
160
+ if (existsSync(join(cwd, candidate))) {
161
+ const candExt = extname(candidate)
162
+ if (candExt === '.py') {
163
+ return {
164
+ command: `pytest ${candidate} -q --tb=short`,
165
+ testFile: candidate,
166
+ type: 'python-slice',
167
+ selection: `slice for ${normalized} -> ${candidate}`,
168
+ }
169
+ }
170
+ if (candExt === '.ts' || candExt === '.tsx') {
171
+ return {
172
+ command: `npx tsx ${candidate}`,
173
+ testFile: candidate,
174
+ type: 'ts-slice',
175
+ selection: `slice for ${normalized} -> ${candidate}`,
176
+ }
177
+ }
178
+ return {
179
+ command: `node --test ${candidate}`,
180
+ testFile: candidate,
181
+ type: 'node-slice',
182
+ selection: `slice for ${normalized} -> ${candidate}`,
183
+ }
184
+ }
185
+ }
186
+
187
+ // 3. Mapping gap: no direct test file found
188
+ return {
189
+ command: null,
190
+ testFile: null,
191
+ type: 'mapping-gap',
192
+ fallback: `npx biome check ${normalized}`,
193
+ selection: `mapping gap for ${normalized}: no direct test file found; use static check fallback`,
194
+ }
195
+ }
196
+
197
+ /**
198
+ * Determine the optimal focused test slice for a set of changed files.
199
+ */
200
+ export function resolveTestSlice(files = [], options = {}) {
201
+ const cwd = options.cwd || process.cwd()
202
+ const failedOnly = options.failedOnly
203
+
204
+ // Filter for relevant source or test files
205
+ const relevantFiles = files.filter((f) => {
206
+ const ext = extname(f)
207
+ return SOURCE_EXTENSIONS.has(ext) || TEST_EXTENSIONS.has(ext) || f.includes('.test.')
208
+ })
209
+
210
+ // 1. If repo has a predefined focused script and multiple files changed, prefer it
211
+ const repoScript = findRepoSliceScript(cwd)
212
+ if (repoScript && (relevantFiles.length === 0 || relevantFiles.length > 2)) {
213
+ return {
214
+ command: repoScript.command,
215
+ type: 'repo-selector',
216
+ selection: `repo selector: ${repoScript.name}`,
217
+ files: relevantFiles,
218
+ }
219
+ }
220
+
221
+ // 2. If 1 or 2 specific files changed, map to the direct slice
222
+ if (relevantFiles.length > 0) {
223
+ const primaryFile = relevantFiles[0]
224
+ const slice = mapFileToSlice(primaryFile, cwd)
225
+
226
+ if (slice.command) {
227
+ let finalCommand = slice.command
228
+ if (failedOnly) {
229
+ if (slice.type.startsWith('python')) {
230
+ finalCommand = `pytest ${slice.testFile} -k "${failedOnly}" -q --tb=short`
231
+ } else if (slice.type.startsWith('node')) {
232
+ finalCommand = `node --test --test-name-pattern="${failedOnly}" ${slice.testFile}`
233
+ }
234
+ }
235
+ return {
236
+ command: finalCommand,
237
+ type: slice.type,
238
+ testFile: slice.testFile,
239
+ selection: slice.selection,
240
+ files: relevantFiles,
241
+ }
242
+ }
243
+
244
+ // Mapping gap: fallback to repo selector if present
245
+ if (repoScript) {
246
+ return {
247
+ command: repoScript.command,
248
+ type: 'repo-selector-fallback',
249
+ selection: `mapping gap for ${primaryFile}; fallback to repo selector ${repoScript.name}`,
250
+ files: relevantFiles,
251
+ }
252
+ }
253
+
254
+ return {
255
+ command: slice.fallback,
256
+ type: 'mapping-gap',
257
+ selection: slice.selection,
258
+ files: relevantFiles,
259
+ }
260
+ }
261
+
262
+ // 3. Fallback to repo selector if available
263
+ if (repoScript) {
264
+ return {
265
+ command: repoScript.command,
266
+ type: 'repo-selector',
267
+ selection: `repo selector: ${repoScript.name}`,
268
+ files: [],
269
+ }
270
+ }
271
+
272
+ return {
273
+ command: 'git status',
274
+ type: 'none',
275
+ selection: 'no changed source files detected',
276
+ files: [],
277
+ }
278
+ }
279
+
280
+ /**
281
+ * Execute the slice and produce an observable summary receipt.
282
+ */
283
+ export function executeTestSlice(slicePlan, options = {}) {
284
+ const cwd = options.cwd || process.cwd()
285
+ const startTime = Date.now()
286
+
287
+ const result = spawnSync(slicePlan.command, {
288
+ cwd,
289
+ shell: true,
290
+ encoding: 'utf8',
291
+ })
292
+
293
+ const durationMs = Date.now() - startTime
294
+ const durationSec = (durationMs / 1000).toFixed(2)
295
+ const isPass = result.status === 0
296
+
297
+ // Count tests from output lines
298
+ const output = `${result.stdout || ''}\n${result.stderr || ''}`
299
+ const passLines = output.split('\n').filter((line) => /✔|PASS|passed/i.test(line))
300
+ const targets = Math.max(1, passLines.length)
301
+
302
+ const receipt = {
303
+ status: isPass ? 'PASS' : 'FAIL',
304
+ command: slicePlan.command,
305
+ targets,
306
+ durationSec,
307
+ durationMs,
308
+ selection: slicePlan.selection,
309
+ exitCode: result.status,
310
+ output: isPass ? output.trim().split('\n').slice(-5).join('\n') : output,
311
+ }
312
+
313
+ const summaryLine = `**tests:** [${receipt.status}] \`${receipt.command}\` (targets: ${receipt.targets}, duration: ${receipt.durationSec}s, selection: ${receipt.selection})`
314
+
315
+ return { ...receipt, summaryLine }
316
+ }
317
+
318
+ // ─── CLI Entrypoint ──────────────────────────────────────────────────────────
319
+
320
+ if (process.argv[1] && import.meta.url.endsWith(basename(process.argv[1]))) {
321
+ const args = process.argv.slice(2)
322
+ const jsonMode = args.includes('--json')
323
+ const runMode = args.includes('--run')
324
+ const failedIdx = args.indexOf('--failed-only')
325
+ const failedOnly = failedIdx !== -1 ? args[failedIdx + 1] : undefined
326
+
327
+ const explicitFiles = args.filter((a) => !a.startsWith('--') && a !== failedOnly)
328
+ const files = explicitFiles.length > 0 ? explicitFiles : getChangedFiles()
329
+
330
+ const plan = resolveTestSlice(files, { failedOnly })
331
+
332
+ if (runMode) {
333
+ const receipt = executeTestSlice(plan)
334
+ if (jsonMode) {
335
+ console.log(JSON.stringify(receipt, null, 2))
336
+ } else {
337
+ console.log(receipt.summaryLine)
338
+ if (receipt.exitCode !== 0) {
339
+ console.error(receipt.output)
340
+ process.exit(receipt.exitCode)
341
+ }
342
+ }
343
+ } else {
344
+ if (jsonMode) {
345
+ console.log(JSON.stringify(plan, null, 2))
346
+ } else {
347
+ console.log(`[Test Slice] Command: ${plan.command}`)
348
+ console.log(`[Test Slice] Selection: ${plan.selection}`)
349
+ console.log(`[Test Slice] Files: ${plan.files.join(', ') || 'none'}`)
350
+ }
351
+ }
352
+ }
package/skills-lock.json CHANGED
@@ -34,7 +34,7 @@
34
34
  "src/skills/tdd-with-agents/references/advanced-testing-patterns.md": "4aac82748bb670a61d67732168f8a265d5e66895a9183029e3da03de8e024788",
35
35
  "src/skills/tdd-with-agents/references/testing-by-layer.md": "545291301661e57d846b2db2acca1edb1219a102dce4aad7e0c3cc1a250b570a",
36
36
  "src/skills/tdd-with-agents/SKILL.md": "107495b0473a561dd570d2c25420329d68712776524a736ccc415171f7aa6c00",
37
- "src/skills/test-slicing/SKILL.md": "ff70cddc66261049250b82f7aeabbfd6bcd00f9abe34cc8d5f20aa35333b1590",
37
+ "src/skills/test-slicing/SKILL.md": "7361f197ab6e59a55003eb571c294f1a07388ee1d3f96b1dbfb69a84276b1bb5",
38
38
  "src/skills/verification-planning/SKILL.md": "6a3c6cb5bd5076862d369885bf0b3ba086b625495fe7a42f3f5116d07b470f3e",
39
39
  "src/skills/visual-review-pipeline/references/findings-schema.md": "776511757c0d8387fab232fced8ac9ce031a03f62b6ca539e3c38b0d45035c8e",
40
40
  "src/skills/visual-review-pipeline/references/workflow-steps.md": "13796a32076dd4d04d37a50927169929ba44a332302bd98f72ec61482ceedc09",
@@ -3,6 +3,7 @@ name: aphrodite
3
3
  description: "Frontend specialist — React 19, TypeScript strict, WCAG accessibility, responsive design, TDD, modern API patterns, deprecated npm detection. Routes discovery and Themis review requests through Zeus."
4
4
  mode: all
5
5
  reasoning_effort: medium
6
+ temperature: 0.3
6
7
  permission:
7
8
  read: allow
8
9
  grep: allow
@@ -15,7 +16,13 @@ permission:
15
16
  question: allow
16
17
  task:
17
18
  "*": deny
18
- temperature: 0.3
19
+ todowrite: allow
20
+ todoread: allow
21
+ mcp_tools:
22
+ pantheon-resources: all
23
+ pantheon-memory: [memory_search]
24
+ pantheon-code-mode: [execute_code_script]
25
+ pantheon-vision: [vision_describe, vision_ocr, vision_analyze]
19
26
  skills:
20
27
  - fast-path-execution
21
28
  - incremental-implementation
@@ -26,11 +33,6 @@ skills:
26
33
  - test-slicing
27
34
  - visual-review-pipeline
28
35
  - worktrees
29
- mcp_tools:
30
- pantheon-resources: all
31
- pantheon-memory: [memory_search]
32
- pantheon-code-mode: [execute_code_script]
33
- pantheon-vision: [vision_describe, vision_ocr, vision_analyze]
34
36
  ---
35
37
 
36
38
  ## MCP governance (identical ladder, grant-aware)
@@ -2,10 +2,8 @@
2
2
  name: apollo
3
3
  description: "Read-only investigation scout — broad multi-query discovery when independent searches materially reduce uncertainty. Called by: athena, zeus, hermes, aphrodite, demeter. No edits, no commands."
4
4
  mode: subagent
5
- visible: false
6
- tools:
7
- task: false
8
5
  reasoning_effort: low
6
+ temperature: 0.1
9
7
  permission:
10
8
  read: allow
11
9
  grep: allow
@@ -17,7 +15,14 @@ permission:
17
15
  bash: deny
18
16
  task:
19
17
  "*": deny
20
- temperature: 0.1
18
+ todowrite: allow
19
+ todoread: allow
20
+ mcp_tools:
21
+ pantheon-resources: all
22
+ pantheon-memory:
23
+ - memory_search
24
+ - code_query
25
+ - code_neighbors
21
26
  skills:
22
27
  - auto-continue
23
28
  - codemap
@@ -25,9 +30,6 @@ skills:
25
30
  - reflect
26
31
  - simplify
27
32
  - worktrees
28
- mcp_tools:
29
- pantheon-resources: all
30
- pantheon-memory: [memory_search, code_query, code_neighbors]
31
33
  ---
32
34
 
33
35
  ## When NOT to Use Apollo
@@ -1,22 +1,8 @@
1
1
  ---
2
2
  name: athena
3
- description: Strategic planner & architect — research-first, plan-only, never implements.
4
- Plans include quality gates (ruff/Biome, dep detection, LTS policy). Root Athena
5
- may call Apollo for discovery; child sessions do not delegate.
3
+ description: "Strategic planner & architect — research-first, plan-only, never implements. Plans include quality gates (ruff/Biome, dep detection, LTS policy). Root Athena may call Apollo for discovery; child sessions do not delegate."
6
4
  mode: all
7
5
  reasoning_effort: medium
8
-
9
- mcp_tools:
10
- pantheon-resources: all
11
- pantheon-memory: [memory_search, code_query, code_neighbors]
12
- skills:
13
- - loop-engineering
14
- - memory-bank
15
- - reflect
16
- - simplify
17
- - spec-driven-development
18
- - verification-planning
19
- - worktrees
20
6
  permission:
21
7
  edit: deny
22
8
  hashline_edit: deny
@@ -28,9 +14,25 @@ permission:
28
14
  webfetch: allow
29
15
  pantheon_deepwork: allow
30
16
  question: allow
17
+ todowrite: allow
31
18
  task:
32
19
  "*": deny
33
20
  apollo: allow
21
+ todoread: allow
22
+ mcp_tools:
23
+ pantheon-resources: all
24
+ pantheon-memory:
25
+ - memory_search
26
+ - code_query
27
+ - code_neighbors
28
+ skills:
29
+ - loop-engineering
30
+ - memory-bank
31
+ - reflect
32
+ - simplify
33
+ - spec-driven-development
34
+ - verification-planning
35
+ - worktrees
34
36
  ---
35
37
 
36
38
  ## MCP governance (identical ladder, grant-aware)
@@ -76,9 +78,10 @@ Only an Athena root session may delegate discovery to @apollo. A child session h
76
78
  1. **Understand** the user's goal and requirements
77
79
  2. **Research** the codebase directly; if Athena is root and broader discovery is material, delegate it to @apollo
78
80
  3. **Plan** in detailed, actionable phases (3-5 max, not 10+)
79
- 4. **Return the complete plan** to the caller using the Plan Handoff contract below
80
- 5. **Ask for approval only when directly invoked**; if delegated, Zeus owns the user-facing gate
81
- 6. **Handoff** the approved scope to @zeus when the user also requested execution
81
+ 4. **Mirror to TUI Todo block:** Call `todowrite` with the sequential phases (status `pending`) so the active session displays the plan in the TUI Todo sidebar
82
+ 5. **Return the complete plan** to the caller using the Plan Handoff contract below
83
+ 6. **Ask for approval only when directly invoked**; if delegated, Zeus owns the user-facing gate
84
+ 7. **Handoff** the approved scope to @zeus when the user also requested execution
82
85
 
83
86
  ## Model Source of Truth
84
87
 
@@ -3,6 +3,7 @@ name: demeter
3
3
  description: "Database specialist — SQLAlchemy 2.0, Alembic, query optimization, N+1 prevention, TDD migrations, modern DB libs. Requests Zeus to route broad discovery to Apollo; sends review needs to Zeus."
4
4
  mode: all
5
5
  reasoning_effort: medium
6
+ temperature: 0.2
6
7
  permission:
7
8
  read: allow
8
9
  grep: allow
@@ -13,7 +14,12 @@ permission:
13
14
  webfetch: allow
14
15
  task:
15
16
  "*": deny
16
- temperature: 0.2
17
+ todowrite: allow
18
+ todoread: allow
19
+ mcp_tools:
20
+ pantheon-resources: all
21
+ pantheon-memory: [memory_search]
22
+ pantheon-code-mode: [execute_code_script]
17
23
  skills:
18
24
  - incremental-implementation
19
25
  - loop-engineering
@@ -21,10 +27,6 @@ skills:
21
27
  - simplify
22
28
  - tdd-with-agents
23
29
  - worktrees
24
- mcp_tools:
25
- pantheon-resources: all
26
- pantheon-memory: [memory_search]
27
- pantheon-code-mode: [execute_code_script]
28
30
  ---
29
31
 
30
32
  ## MCP governance (identical ladder, grant-aware)
@@ -2,10 +2,8 @@
2
2
  name: gaia
3
3
  description: "Remote sensing domain specialist — satellite image processing, spectral analysis, SAR, change detection, time series, ML/DL classification. Read-only analysis of geospatial data."
4
4
  mode: subagent
5
- visible: false
6
- tools:
7
- task: false
8
5
  reasoning_effort: low
6
+ temperature: 0.2
9
7
  permission:
10
8
  read: allow
11
9
  grep: allow
@@ -16,17 +14,18 @@ permission:
16
14
  bash: deny
17
15
  task:
18
16
  "*": deny
19
- temperature: 0.2
17
+ todowrite: allow
18
+ todoread: allow
19
+ mcp_tools:
20
+ pantheon-resources: all
21
+ pantheon-memory: [memory_recall]
22
+ pantheon-vision: [vision_describe, vision_ocr, vision_analyze]
20
23
  skills:
21
24
  - auto-continue
22
25
  - loop-engineering
23
26
  - reflect
24
27
  - simplify
25
28
  - worktrees
26
- mcp_tools:
27
- pantheon-resources: all
28
- pantheon-memory: [memory_recall]
29
- pantheon-vision: [vision_describe, vision_ocr, vision_analyze]
30
29
  ---
31
30
 
32
31
  ## MCP governance (identical ladder, grant-aware)
@@ -3,6 +3,7 @@ name: hephaestus
3
3
  description: "AI tooling & pipelines specialist — LangChain/LangGraph chains, RAG architecture, vector stores, embedding strategies. Forges AI infrastructure. Requests Zeus to route broad discovery to Apollo and review to Themis."
4
4
  mode: all
5
5
  reasoning_effort: medium
6
+ temperature: 0.3
6
7
  permission:
7
8
  read: allow
8
9
  grep: allow
@@ -12,7 +13,12 @@ permission:
12
13
  webfetch: allow
13
14
  task:
14
15
  "*": deny
15
- temperature: 0.3
16
+ todowrite: allow
17
+ todoread: allow
18
+ mcp_tools:
19
+ pantheon-resources: all
20
+ pantheon-memory: [memory_search]
21
+ pantheon-code-mode: [execute_code_script]
16
22
  skills:
17
23
  - auto-continue
18
24
  - clonedeps
@@ -21,10 +27,6 @@ skills:
21
27
  - simplify
22
28
  - tdd-with-agents
23
29
  - worktrees
24
- mcp_tools:
25
- pantheon-resources: all
26
- pantheon-memory: [memory_search]
27
- pantheon-code-mode: [execute_code_script]
28
30
  ---
29
31
 
30
32
  ## MCP governance (identical ladder, grant-aware)
@@ -1,16 +1,26 @@
1
1
  ---
2
2
  name: hermes
3
- description: Backend specialist — FastAPI, Python, async, modern Python stdlib,
4
- obsolete lib detection. Root Hermes may call Apollo for discovery; child
5
- sessions do not delegate. Sends material work to Themis via Zeus.
3
+ description: "Backend specialist — FastAPI, Python, async, modern Python stdlib, obsolete lib detection. Root Hermes may call Apollo for discovery; child sessions do not delegate. Sends material work to Themis via Zeus."
6
4
  mode: all
7
5
  reasoning_effort: medium
8
-
6
+ temperature: 0.3
7
+ permission:
8
+ bash: allow
9
+ shell: allow
10
+ read: allow
11
+ grep: allow
12
+ edit: allow
13
+ webfetch: allow
14
+ glob: allow
15
+ task:
16
+ "*": deny
17
+ apollo: allow
18
+ todowrite: allow
19
+ todoread: allow
9
20
  mcp_tools:
10
21
  pantheon-resources: all
11
22
  pantheon-memory: [memory_search]
12
23
  pantheon-code-mode: [execute_code_script]
13
- temperature: 0.3
14
24
  skills:
15
25
  - clonedeps
16
26
  - fast-path-execution
@@ -22,19 +32,13 @@ skills:
22
32
  - tdd-with-agents
23
33
  - test-slicing
24
34
  - worktrees
25
- permission:
26
- bash: allow
27
- shell: allow
28
- read: allow
29
- grep: allow
30
- edit: allow
31
- webfetch: allow
32
- glob: allow
33
- task:
34
- "*": deny
35
- apollo: allow
36
35
  ---
37
36
 
37
+ ## Strict Test Slicing Mandate
38
+
39
+ - **Targeted Slices Only:** In TDD loops, run ONLY the test slice for the changed file (`pytest tests/test_<file>.py -k <func>`, `node --test tests/<file>.test.mjs`).
40
+ - **NEVER RUN `npm test`:** Unconstrained suites destroy execution speed and blow context windows.
41
+
38
42
  # Hermes - Backend Executor (FastAPI Specialist)
39
43
 
40
44
  ## When NOT to Use Hermes
@@ -3,21 +3,26 @@ name: iris
3
3
  description: "GitHub operations specialist — branches, pull requests, issues, releases, tags. Called by zeus after review. Never pushes or merges without explicit human approval. Integrates with GitHub CLI (gh) for operations."
4
4
  mode: all
5
5
  reasoning_effort: low
6
+ temperature: 0.2
6
7
  permission:
7
8
  read: allow
8
9
  grep: allow
9
10
  bash:
10
- git *: allow
11
- gh *: allow
11
+ "git *": allow
12
+ "gh *": allow
12
13
  shell:
13
- git *: allow
14
- gh *: allow
14
+ "git *": allow
15
+ "gh *": allow
15
16
  webfetch: allow
16
17
  edit: deny
17
18
  hashline_edit: deny
18
19
  task:
19
20
  "*": deny
20
- temperature: 0.2
21
+ todowrite: allow
22
+ todoread: allow
23
+ mcp_tools:
24
+ pantheon-resources: all
25
+ pantheon-memory: [memory_recall]
21
26
  skills:
22
27
  - artifact-management
23
28
  - git-workflow-and-versioning
@@ -25,9 +30,6 @@ skills:
25
30
  - reflect
26
31
  - simplify
27
32
  - worktrees
28
- mcp_tools:
29
- pantheon-resources: all
30
- pantheon-memory: [memory_recall]
31
33
  ---
32
34
 
33
35
  ## MCP governance (identical ladder, grant-aware)