Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commite65f52e

Browse files
committed
tool changes
1 parent8c306ae commite65f52e

File tree

3 files changed

+1
-259
lines changed

3 files changed

+1
-259
lines changed

‎pkg/github/instructions.go‎

Lines changed: 0 additions & 73 deletions
Original file line numberDiff line numberDiff line change
@@ -62,79 +62,6 @@ Check 'list_issue_types' first for organizations to use proper issue types. Use
6262
return`## Discussions
6363
6464
Use 'list_discussion_categories' to understand available categories before creating discussions. Filter by category for better organization.`
65-
case"projects":
66-
return`## Projects
67-
68-
Workflow: 1) list_project_fields (get field IDs), 2) list_project_items (with pagination), 3) optional updates.
69-
70-
Field usage:
71-
- Call list_project_fields first to understand available fields and get IDs/types before filtering.
72-
- Use EXACT returned field names (case-insensitive match). Don't invent names or IDs.
73-
- Iteration synonyms (sprint/cycle) only if that field exists; map to the actual name (e.g. sprint:@current).
74-
- Only include filters for fields that exist and are relevant.
75-
76-
Pagination (mandatory):
77-
- Loop while pageInfo.hasNextPage=true using after=pageInfo.nextCursor.
78-
- Keep query, fields, per_page IDENTICAL on every page.
79-
- Use before=pageInfo.prevCursor only when explicitly navigating to a previous page.
80-
- Do not analyze until ALL pages fetched.
81-
82-
Counting rules:
83-
- Count items array length after full pagination.
84-
- If multi-page: collect all pages, dedupe by item.id (fallback node_id) before totals.
85-
- Never count field objects, content, or nested arrays as separate items.
86-
- item.id = project item ID (for updates/deletes). item.content.id = underlying issue/PR ID.
87-
88-
Summary vs list:
89-
- Summaries ONLY if user uses verbs: analyze | summarize | summary | report | overview | insights.
90-
- Listing verbs (list/show/get/fetch/display/enumerate) → enumerate + total.
91-
92-
Self-check before returning:
93-
- Paginated fully
94-
- Dedupe by id/node_id
95-
- Correct IDs used
96-
- Field names valid
97-
- Summary only if requested.
98-
99-
Return COMPLETE data or state what's missing (e.g. pages skipped).
100-
101-
list_project_items query rules:
102-
Query string - For advanced filtering of project items using GitHub's project filtering syntax:
103-
104-
MUST reflect user intent; strongly prefer explicit content type if narrowed:
105-
- "open issues" → state:open is:issue
106-
- "merged PRs" → state:merged is:pr
107-
- "items updated this week" → updated:>@today-7d (omit type only if mixed desired)
108-
- "list all P1 priority items" → priority:p1 (omit state if user wants all, omit type if user specifies "items")
109-
- "list all open P2 issues" → is:issue state:open priority:p2 (include state if user wants open or closed, include type if user specifies "issues" or "PRs")
110-
- "all open issues I'm working on" → is:issue state:open assignee:@me
111-
112-
Query Construction Heuristics:
113-
a. Extract type nouns: issues → is:issue | PRs, Pulls, or Pull Requests → is:pr | tasks/tickets → is:issue (ask if ambiguity)
114-
b. Map temporal phrases: "this week" → updated:>@today-7d
115-
c. Map negations: "excluding wontfix" → -label:wontfix
116-
d. Map priority adjectives: "high/sev1/p1" → priority:high OR priority:p1 (choose based on field presence)
117-
e. When filtering by label, always use wildcard matching to account for cross-repository differences or emojis: (e.g. "bug 🐛" → label:*bug*)
118-
f. When filtering by milestone, always use wildcard matching to account for cross-repository differences: (e.g. "v1.0" → milestone:*v1.0*)
119-
120-
Syntax Essentials (items):
121-
AND: space-separated. (label:bug priority:high).
122-
OR: comma inside one qualifier (label:bug,critical).
123-
NOT: leading '-' (-label:wontfix).
124-
Hyphenate multi-word field names. (team-name:"Backend Team", story-points:>5).
125-
Quote multi-word values. (status:"In Review" team-name:"Backend Team").
126-
Ranges: points:1..3, updated:<@today-30d.
127-
Wildcards: title:*crash*, label:bug*.
128-
Assigned to User: assignee:@me | assignee:username | no:assignee
129-
130-
Common Qualifier Glossary (items):
131-
is:issue | is:pr | state:open|closed|merged | assignee:@me|username | label:NAME | status:VALUE |
132-
priority:p1|high | sprint-name:@current | team-name:"Backend Team" | parent-issue:"org/repo#123" |
133-
updated:>@today-7d | title:*text* | -label:wontfix | label:bug,critical | no:assignee | has:label
134-
135-
Never:
136-
- Infer field IDs; fetch via list_project_fields.
137-
- Drop 'fields' param on subsequent pages if field values are needed.`
13865
default:
13966
return""
14067
}

‎pkg/github/projects.go‎

Lines changed: 0 additions & 183 deletions
Original file line numberDiff line numberDiff line change
@@ -1066,186 +1066,3 @@ func addOptions(s string, opts any) (string, error) {
10661066
origURL.RawQuery=origValues.Encode()
10671067
returnorigURL.String(),nil
10681068
}
1069-
1070-
funcManageProjectItemsPrompt(t translations.TranslationHelperFunc) (tool mcp.Prompt,handler server.PromptHandlerFunc) {
1071-
returnmcp.NewPrompt("ManageProjectItems",
1072-
mcp.WithPromptDescription(t("PROMPT_MANAGE_PROJECT_ITEMS_DESCRIPTION","Guide for GitHub Projects V2: discovery, fields, querying, updates.")),
1073-
mcp.WithArgument("owner",mcp.ArgumentDescription("The owner of the project (user or organization name)"),mcp.RequiredArgument()),
1074-
mcp.WithArgument("owner_type",mcp.ArgumentDescription("Type of owner: 'user' or 'org'"),mcp.RequiredArgument()),
1075-
),func(_ context.Context,request mcp.GetPromptRequest) (*mcp.GetPromptResult,error) {
1076-
owner:=request.Params.Arguments["owner"]
1077-
ownerType:=request.Params.Arguments["owner_type"]
1078-
1079-
task:=""
1080-
ift,exists:=request.Params.Arguments["task"];exists {
1081-
task=fmt.Sprintf("%v",t)
1082-
}
1083-
1084-
messages:= []mcp.PromptMessage{
1085-
{
1086-
Role:"system",
1087-
Content:mcp.NewTextContent(`System guide: GitHub Projects V2.
1088-
Goal: Pick correct tool, fetch COMPLETE data (no early pagination stop), apply accurate filters, and count items correctly.
1089-
1090-
Core rules:
1091-
- list_projects: NEVER include item-level filters (no is:issue, assignee:, label:, etc.)
1092-
- Before filtering on fields, call list_project_fields to get field IDs
1093-
- Always paginate until pageInfo.hasNextPage=false
1094-
- Keep query, fields, per_page identical across pages
1095-
- Include field IDs on every list_project_items page if you need values
1096-
- Prefer explicit is:issue / is:pr unless mixed set requested
1097-
- Only summarize if verbs like analyze / summarize / report / overview / insights appear; otherwise enumerate
1098-
1099-
Field resolution:
1100-
- Use exact returned field names; don't invent
1101-
- Iteration synonyms map to actual existing name (Sprint → sprint:@current, etc.). If none exist, omit
1102-
- Only add filters for fields that exist and matter to the user goal
1103-
1104-
Query syntax essentials:
1105-
AND space | OR comma | NOT prefix - | quote multi-word values | hyphenate names | ranges points:1..5 | comparisons updated:>@today-7d priority:>1 | wildcards title:*crash*
1106-
1107-
Pagination pattern:
1108-
Call list_project_items → if hasNextPage true, repeat with after=nextCursor → stop only when false → then count/deduplicate
1109-
1110-
Counting:
1111-
- Items array length after full pagination (dedupe by item.id or node_id)
1112-
- Never count fields array, content, assignees, labels as separate items
1113-
- item.id = project item identifier; content.id = underlying issue/PR id
1114-
1115-
Edge handling:
1116-
Empty pages → total=0 still return pageInfo
1117-
Duplicates → keep first for totals
1118-
Missing field values → null/omit, never fabricate
1119-
1120-
Self-check: paginated? deduped? correct IDs? field names valid? summary allowed?`),
1121-
},
1122-
{
1123-
Role:"user",
1124-
Content:mcp.NewTextContent(fmt.Sprintf("I want to work with GitHub Projects for %s (owner_type: %s).%s",
1125-
owner,
1126-
ownerType,
1127-
func()string {
1128-
iftask!="" {
1129-
returnfmt.Sprintf(" Focus: %s.",task)
1130-
}
1131-
return""
1132-
}())),
1133-
},
1134-
{
1135-
Role:"assistant",
1136-
Content:mcp.NewTextContent("Start by listing projects: use list_projects tool with owner and owner_type parameters."),
1137-
},
1138-
{
1139-
Role:"user",
1140-
Content:mcp.NewTextContent("How do I work with fields and items?"),
1141-
},
1142-
{
1143-
Role:"assistant",
1144-
Content:mcp.NewTextContent(`Fields & items workflow:
1145-
1. Call list_project_fields to get field definitions → map lowercased name -> {id,type}
1146-
2. Use only existing field names; no invention
1147-
3. Iteration mapping: pick sprint/cycle/iteration only if present (sprint:@current etc.)
1148-
4. Include only relevant fields (e.g. Priority + Label for high priority bugs)
1149-
5. Build query after resolving fields ("last week" → updated:>@today-7d)
1150-
6. Call list_project_items with query and field IDs → paginate until hasNextPage=false
1151-
7. Keep query/fields/per_page stable across all pages
1152-
8. Include field IDs on every page when you need their values
1153-
Missing field? Omit or clarify—never guess.`),
1154-
},
1155-
{
1156-
Role:"user",
1157-
Content:mcp.NewTextContent("How do I update item field values?"),
1158-
},
1159-
{
1160-
Role:"assistant",
1161-
Content:mcp.NewTextContent(`Updating fields (update_project_item tool):
1162-
Input format: updated_field parameter with {id: <field_id>, value: <new_value>}
1163-
Examples:
1164-
- Text field: {"id":123,"value":"hello"}
1165-
- Single-select: {"id":456,"value":789} (value is option ID, not name)
1166-
- Number: {"id":321,"value":5}
1167-
- Date: {"id":654,"value":"2025-03-15"}
1168-
- Clear field: {"id":123,"value":null}
1169-
1170-
Rules:
1171-
- item_id parameter = project item ID (from list_project_items), NOT issue/PR ID
1172-
- Get field IDs from list_project_fields first
1173-
- For select/iteration fields, pass option/iteration ID as value, not the name
1174-
- To add an item first: use add_project_item tool with issue/PR ID
1175-
- To remove an item: use delete_project_item tool`),
1176-
},
1177-
{
1178-
Role:"user",
1179-
Content:mcp.NewTextContent("Show me a workflow example."),
1180-
},
1181-
{
1182-
Role:"assistant",
1183-
Content:mcp.NewTextContent(`Workflow example:
1184-
1. list_projects → pick project_number
1185-
2. list_project_fields → build field map {name: {id, type}}
1186-
3. Build query (e.g. is:issue sprint:@current priority:high updated:>@today-7d)
1187-
4. list_project_items with field IDs → paginate fully (loop until hasNextPage=false)
1188-
5. Optional: get_project_item for specific item details
1189-
6. Optional: add_project_item to add issue/PR to project
1190-
7. Optional: update_project_item to change field values for an item
1191-
8. Optional: delete_project_item to remove from an item from project
1192-
1193-
Important:
1194-
- Iteration filter must match existing field name
1195-
- Keep fields parameter consistent across pages
1196-
- Summarize only if explicitly asked
1197-
- item_id for updates/deletes comes from list_project_items response
1198-
- content.id in items is the underlying issue/PR ID (use with add_project_item)`),
1199-
},
1200-
{
1201-
Role:"user",
1202-
Content:mcp.NewTextContent("How do I handle pagination?"),
1203-
},
1204-
{
1205-
Role:"assistant",
1206-
Content:mcp.NewTextContent(`Pagination with list_project_items:
1207-
1. Make initial call with query, fields, per_page parameters
1208-
2. Check response pageInfo.hasNextPage
1209-
3. If true: call again with same query/fields/per_page + after=pageInfo.nextCursor
1210-
4. Repeat step 2-3 until hasNextPage=false
1211-
5. Collect all items from all pages before counting/analyzing
1212-
1213-
Critical: Do NOT change query, fields, or per_page between pages. Always include same field IDs on every page if you need field values.`),
1214-
},
1215-
{
1216-
Role:"user",
1217-
Content:mcp.NewTextContent("How do I get more details about items?"),
1218-
},
1219-
{
1220-
Role:"assistant",
1221-
Content:mcp.NewTextContent(`Getting additional item details:
1222-
1223-
- First inspect item's content object for info, e.g. title, assignees, labels
1224-
- If additional detail is needed, and relevant fields are present from list_project_fields, include their IDs in list_project_items and request with list_project_items again.
1225-
- If more detail needed, use separate issue/PR tools`),
1226-
},
1227-
{
1228-
Role:"assistant",
1229-
Content:mcp.NewTextContent(`Query patterns for list_project_items:
1230-
1231-
Common scenarios:
1232-
- Blocked issues: is:issue (label:blocked OR status:"Blocked")
1233-
- Overdue tasks: is:issue due-date:<@today state:open
1234-
- PRs ready for review: is:pr review-status:"Ready for Review" state:open
1235-
- Stale issues: is:issue updated:<@today-30d state:open
1236-
- High priority bugs: is:issue label:bug priority:high state:open
1237-
- Team sprint PRs: is:pr team-name:"Backend Team" sprint:@current
1238-
1239-
Rules:
1240-
- Summarize only if user requests it with verbs like "analyze", "summarize", "report"
1241-
- Deduplicate by item.id before counting totals
1242-
- Quote multi-word values: status:"In Progress"
1243-
- Never invent field names or IDs - always verify with list_project_fields first
1244-
- Use explicit is:issue or is:pr unless user wants mixed items`),
1245-
},
1246-
}
1247-
return&mcp.GetPromptResult{
1248-
Messages:messages,
1249-
},nil
1250-
}
1251-
}

‎pkg/github/tools.go‎

Lines changed: 1 addition & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -328,9 +328,7 @@ func DefaultToolsetGroup(readOnly bool, getClient GetClientFn, getGQLClient GetG
328328
toolsets.NewServerTool(AddProjectItem(getClient,t)),
329329
toolsets.NewServerTool(DeleteProjectItem(getClient,t)),
330330
toolsets.NewServerTool(UpdateProjectItem(getClient,t)),
331-
).AddPrompts(
332-
toolsets.NewServerPrompt(ManageProjectItemsPrompt(t)),
333-
)
331+
)
334332
stargazers:=toolsets.NewToolset(ToolsetMetadataStargazers.ID,ToolsetMetadataStargazers.Description).
335333
AddReadTools(
336334
toolsets.NewServerTool(ListStarredRepositories(getClient,t)),

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp