pipebrain-mcp 0.47.0 → 0.47.1

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 (2) hide show
  1. package/dist/main.js +3 -3
  2. package/package.json +1 -1
package/dist/main.js CHANGED
@@ -17771,7 +17771,7 @@ function createMcpServer(baseClient, apiBaseUrl, orgInstructions2 = /* @__PURE__
17771
17771
  // one description loaded into EVERY session by `anthropic/alwaysLoad`. The clauses
17772
17772
  // that went were the ones arguing for the rule rather than stating it; what a caller
17773
17773
  // needs to predict a refusal all stayed.
17774
- description: `Everything a session needs before it starts work, in ONE read \u2014 assembled by the API and returned verbatim, with the BINDING GUIDANCE appended after it: the product doctrine and this organization\u2019s own writing rules, in full. This is the one call that returns them; every other read only names it. CALL IT FIRST, before you write anything. Addressed by \`issue\` OR by \`project\`, never both and never neither: an issue names its own project, so the pair is refused rather than resolved for you (\xA77). Given an ISSUE: that issue\u2019s full detail \u2014 the same payload \`get_issue\` returns \u2014 PLUS the BODY TEXT of every requirement it links, at the revision the link is pinned at. That text is the round trip this removes: \`get_issue\` names which documents bind the work and at which revision, never what they SAY. Comments come CAPPED at the newest \`commentLimit\` (default ${WORK_CONTEXT_COMMENT_LIMIT}), with \`commentCount\` \u2014 the issue\u2019s TRUE total \u2014 beside them; read the issue directly for the rest. Given a PROJECT: \`available\`, the work a session could START right now. It EXCLUDES finished and abandoned work (\`done\`, \`killed\`), work another person has claimed, anything blocked \u2014 the status guard would refuse to start that \u2014 and work an operator has reserved for a person, which no token may claim however it is reached. Your OWN claims stay, and so does unscheduled work carrying no milestone. \`available\` comes CAPPED at \`availableLimit\` rows, with \`availableCount\` \u2014 the project\u2019s TRUE total of startable work \u2014 beside it. The item form carries no \`available\`; the project form carries no item. BOTH forms also carry the project half: the release currently being worked (the active milestone) and the project\u2019s cross-cutting requirements WITH their text \u2014 they bind every epic automatically \u2014 and this organization\u2019s approved conventions in full, standing guidance that binds work in EVERY project here, not just this one. READ-ONLY \u2014 this claims nothing, moves nothing and writes nothing.`,
17774
+ description: `Everything a session needs before it starts work, in ONE read \u2014 assembled by the API and returned verbatim, with the BINDING GUIDANCE appended after it: the product doctrine and this organization\u2019s own writing rules, in full. This is the one call that returns them; every other read only names it. CALL IT FIRST, before you write anything. Addressed by \`issue\` OR by \`project\`, never both and never neither: an issue names its own project, so the pair is refused rather than resolved for you (\xA77). Given an ISSUE: that issue\u2019s full detail \u2014 the same payload \`get_issue\` returns \u2014 PLUS the BODY TEXT of every requirement it links, at the revision the link is pinned at. That text is the round trip this removes: \`get_issue\` names which documents bind the work and at which revision, never what they SAY. Comments come CAPPED at the newest \`commentLimit\` (default ${WORK_CONTEXT_COMMENT_LIMIT}), with \`commentCount\` \u2014 the issue\u2019s TRUE total \u2014 beside them; read the issue directly for the rest. Given a PROJECT: \`available\`, the work a session could START right now. It EXCLUDES finished and abandoned work (\`done\`, \`killed\`), work another person has claimed, anything blocked \u2014 the status guard would refuse to start that \u2014 and work an operator has reserved for a person, which no token may claim \u2014 a reservation refuses the claim and nothing else. Your OWN claims stay, and so does unscheduled work carrying no milestone. \`available\` comes CAPPED at \`availableLimit\` rows, with \`availableCount\` \u2014 the project\u2019s TRUE total of startable work \u2014 beside it. The item form carries no \`available\`; the project form carries no item. BOTH forms also carry the project half: the release currently being worked (the active milestone) and the project\u2019s cross-cutting requirements WITH their text \u2014 they bind every epic automatically \u2014 and this organization\u2019s approved conventions in full, standing guidance that binds work in EVERY project here, not just this one. READ-ONLY \u2014 this claims nothing, moves nothing and writes nothing.`,
17775
17775
  // The WHOLE domain schema rather than its `.shape`: the exactly-one-of rule is a
17776
17776
  // refinement on the OBJECT, and a raw shape drops it — the SDK validates against
17777
17777
  // what it is handed, so the shape alone would quietly accept both arguments and
@@ -17964,7 +17964,7 @@ function createMcpServer(baseClient, apiBaseUrl, orgInstructions2 = /* @__PURE__
17964
17964
  inputSchema: setStatusInputSchema.shape
17965
17965
  }, (args) => run(() => client2().setStatus(args.issue, args.status)));
17966
17966
  registerStrict("claim_issue", {
17967
- description: "Atomically claim an issue (\xA76 rule 1): succeeds only if unclaimed; under a race exactly one wins and the loser\u2019s error names the current claimer. Claim is orthogonal to status (\xA77). REFUSED with `reserved_for_person` when an operator has reserved the work for a person: reading and commenting stay open to you, and only the claim is refused. Clearing that mark is a browser-session act, so there is no call here that lifts it \u2014 say in a comment why the work looks agent-eligible and leave the choice to the operator.",
17967
+ description: "Atomically claim an issue (\xA76 rule 1): succeeds only if unclaimed; under a race exactly one wins and the loser\u2019s error names the current claimer. Claim is orthogonal to status (\xA77). REFUSED with `reserved_for_person` when an operator has reserved the work for a person, and THE CLAIM IS THE ONLY CALL A RESERVATION REFUSES \u2014 reading it, commenting on it, moving its status and releasing a claim all still succeed. The mark stops you taking the work, not touching it, so do not read it as a general instruction to leave the item alone. Clearing the mark is a browser-session act, so no call here lifts it: say in a comment why the work looks agent-eligible and leave the choice to the operator.",
17968
17968
  inputSchema: { issue: issueRefSchema }
17969
17969
  }, (args) => run(() => client2().claimIssue(args.issue)));
17970
17970
  registerStrict("release_issue", {
@@ -18341,7 +18341,7 @@ function loadMcpConfig() {
18341
18341
  }
18342
18342
 
18343
18343
  // src/version.ts
18344
- var VERSION = "0.47.0";
18344
+ var VERSION = "0.47.1";
18345
18345
 
18346
18346
  // src/main.ts
18347
18347
  async function scopedClient(base, org) {
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "pipebrain-mcp",
3
- "version": "0.47.0",
3
+ "version": "0.47.1",
4
4
  "description": "Pipebrain MCP server for Claude Code and Claude Desktop — issues, requirements, decision records, conventions, prompts, content, assets, metadata-only secrets, and more of the platform's modules over the Pipebrain API.",
5
5
  "type": "module",
6
6
  "license": "MIT",