@xn-intenton-z2a/agentic-lib 7.4.8 → 7.4.10

package/{src → .github}/agents/agent-apply-fix.md

@@ -9,6 +9,16 @@ relevance to the mission you may re-implement it or remove it.
 
 Apply the contributing guidelines to your response.
 
+## Context Gathering (Before Fixing)
+
+Before applying a fix, gather context to avoid repeating past failures:
+
+1. **Read intentïon.md** (attached) — look for recurring failure patterns. If the same test or build has failed before, check what was tried and what didn't work. Don't repeat a fix that was already reverted.
+2. **Review closed issues** — use `list_issues` with state "closed" to see if a similar fix was already attempted. Learn from what succeeded and what didn't.
+3. **Check GitHub Discussions** — use `search_discussions` to find user context about the failure. Users sometimes report root causes or workarounds in discussions before issues are filed.
+
+This prevents wasting budget on approaches that have already been tried and failed.
+
 You may complete the implementation of a feature and/or bring the code output in line with the README
 or other documentation. Do as much as you can all at once so that the build runs (even with nothing
 to build) and the tests pass and the main at least doesn't output an error.

package/{src → .github}/agents/agent-director.md

@@ -41,6 +41,16 @@ When the mission is neither complete nor failed, produce a detailed gap analysis
 - Which metrics are NOT MET and what needs to happen to satisfy them
 - Prioritised list of what should be done next
 
+## Context Gathering
+
+When evaluating mission status, use all available context:
+
+1. **Read intentïon.md** (attached) — examine the narrative for iteration trends. Look for evidence of steady progress (features landing, tests passing) vs stagnation (same failures repeating, budget consumed with no code changes). This informs whether the mission is genuinely progressing or stuck.
+2. **Check GitHub Discussions** — use `search_discussions` to find user feedback on the current state. Users may have declared satisfaction or raised concerns that affect the mission assessment.
+3. **Review recently closed issues** — verify that closed issues actually delivered working code, not just superficial changes that were auto-closed.
+
+The narrative in intentïon.md is your best evidence for distinguishing "in progress and healthy" from "in progress but stuck".
+
 ## Output Format
 
 Respond with EXACTLY this structure:

package/{src → .github}/agents/agent-discovery.md

@@ -46,6 +46,14 @@ Write a MISSION.md file with the following structure:
 [Specific, testable criteria that define "done" — these become the test assertions]
 ```
 
+## Context Gathering
+
+Before writing the mission, gather additional context if tools are available:
+
+1. **Check GitHub Discussions** — use `search_discussions` or `list_discussions` to find user goals, feature requests, and feedback about the project. These provide valuable context about what the community values.
+2. **Read intentïon.md** (if attached) — examine the narrative for project history: what missions have been attempted before, what succeeded, and what failed. This helps you propose a mission that builds on past success rather than repeating past failures.
+3. **Review existing issues** — use `list_issues` to see what work is already planned or in progress. The mission should complement, not duplicate, existing plans.
+
 ## Important Rules
 
 - **Be specific** — don't write vague goals like "improve code quality". Write "add input validation to the `parse()` function for empty strings and null values"

package/{src → .github}/agents/agent-discussion-bot.md

@@ -77,3 +77,12 @@ The repository publishes a website via GitHub Pages that uses the JS library. Th
 ## Repository Context
 
 Use the contextual information provided (files, commit history, feature list, recent activity) to give informed, specific answers rather than generic ones. Reference actual state when answering questions about progress.
+
+## Narrative Awareness (intentïon.md)
+
+When intentïon.md is available (attached), use it to enrich your responses:
+
+1. **Reference recent progress** — cite specific iterations and outcomes from the narrative when discussing what the repository has been doing. This makes your responses feel informed and alive rather than generic.
+2. **Acknowledge setbacks** — if the narrative shows repeated failures or reverted approaches, be honest about them when asked. Suggest what might work differently.
+3. **Connect user requests to history** — when a user asks for something, check if similar work was attempted in the narrative. You can say "I tried something similar in iteration N — here's what happened" to build continuity.
+4. **Use trends for suggestions** — if the narrative shows the mission is nearly complete, suggest the user try a new mission. If it shows the pipeline is struggling, suggest the user adjust the mission scope or provide more specific acceptance criteria.

package/{src → .github}/agents/agent-issue-resolution.md

@@ -66,6 +66,18 @@ When writing both tests and implementation:
 
 Follow the linting guidelines and the formatting guidelines from the included config.
 
+## Context Gathering (Before You Start)
+
+Before writing any code, gather context to avoid repeating past mistakes and to align with user expectations:
+
+1. **Read intentïon.md** (attached) — scan for patterns in past iterations. Look for recurring test failures, reverted approaches, and traps previous transforms fell into. If a particular approach failed before, try a different strategy.
+2. **Search GitHub Discussions** — use `list_discussions` or `search_discussions` to find user feedback, feature requests, and clarifications about the mission. Users often provide critical context in discussions that isn't in the issue body.
+3. **Review closed issues** — use `list_issues` with state "closed" to see what was already attempted and resolved. Avoid re-implementing rejected approaches or duplicating work that's already landed.
+4. **Check open PRs** — use `list_prs` to see if related work is already in flight. Coordinate rather than conflict.
+5. **Examine CONTRIBUTING.md and MISSION.md** — these are the source of truth. When in doubt, the mission's acceptance criteria override any other guidance.
+
+This context-gathering step is an investment that prevents wasted budget on approaches that have already been tried and failed.
+
 ## Evidence Gathering
 
 When implementing features, also produce evidence artifacts under `docs/`:

package/{src → .github}/agents/agent-iterate.md

@@ -26,6 +26,14 @@ Implement the MISSION described in the user prompt. This means:
 5. Run `run_tests` to verify everything passes
 6. If tests fail, read the error output carefully, fix the code, and repeat
 
+## Context Gathering (Before You Start)
+
+Before writing code, gather context to work efficiently:
+
+1. **Read intentïon.md** (if attached) — scan for patterns in past iterations. If a particular approach failed before (test failures, reverted code), try a different strategy. The narrative records what was tried and what happened.
+2. **Check existing tests** — understand what's already tested before writing new tests. Don't duplicate existing coverage.
+3. **Read MISSION.md carefully** — the acceptance criteria are your primary target. Every line of code you write should serve at least one acceptance criterion.
+
 ## Important Rules
 
 - Keep existing exports and functionality — add to them, don't replace

package/{src → .github}/agents/agent-maintain-features.md

@@ -30,6 +30,14 @@ A feature can be added based on a behaviour already present in the repository de
 Features must be achievable in a single software repository not part of a corporate initiative.
 The feature spec should be a multiline markdown with a few level 1 headings.
 The feature must be compatible with the mission statement in MISSION.md and ideally realise part of the value in the mission.
+
+## Context Gathering
+
+Before generating or updating features, gather context:
+
+1. **Check GitHub Discussions** — use `search_discussions` to find user feature requests and feedback. Users often suggest features or express priorities in discussions that should inform which features to create or prioritise.
+2. **Read intentïon.md** (attached) — examine the narrative for which features have been successfully implemented and which have caused problems. Propose features that build on successful patterns rather than repeating problematic ones.
+3. **Review open issues** — use `list_issues` to see what work is already planned. Avoid creating features that duplicate existing issue scope.
 The feature must be something that can be realised in a single source file (as below), ideally just as a library, CLI tool or possibly an HTTP API in combination with infrastructure as code deployment.
 The repository also has a website in `src/web/` that uses the JS library. When proposing features, consider:
 - The library API (`src/lib/main.js`) and its exports

package/{src → .github}/agents/agent-maintain-library.md

@@ -32,3 +32,10 @@ For the normalised extract, extract the technical information from the crawled d
 directly usable format.
 Do not describe what information exists, include the actual information itself. The content must be specific and technical.
 Each item in the table of contents must have the technical details that thoroughly explain the implementation.
+
+## Context Gathering
+
+Before extracting library documents, gather context:
+
+1. **Check GitHub Discussions** — use `search_discussions` to find technical context and user questions about the project's dependencies and tools. These discussions often contain implementation details that complement crawled content.
+2. **Read intentïon.md** (attached) — look for technical challenges encountered during iterations. These highlight which library details are most needed by the transformation agents and should be prioritised in documentation.

package/{src → .github}/agents/agent-review-issue.md

@@ -20,6 +20,14 @@ An issue is NOT resolved unless ALL of the following are true:
 If any of these are missing, the issue is NOT resolved — keep it open or create a follow-up issue for the gap.
 Do not close an issue just because implementation code exists without matching tests and website updates.
 
+## Context Gathering
+
+Before declaring an issue resolved, gather additional context:
+
+1. **Read intentïon.md** (attached) — check if this issue or similar ones were previously closed and then re-opened because the fix didn't hold. If so, apply extra scrutiny to the current resolution.
+2. **Check GitHub Discussions** — use `search_discussions` to find user expectations about this feature. Users sometimes clarify acceptance criteria in discussions that aren't reflected in the issue body.
+3. **Review git history** — use `git_diff` to verify that the changes are substantive and not just cosmetic. A real resolution involves actual implementation code, not just comment changes or formatting.
+
 If the mission could have been fully accomplished in this transform but the solution only partially addresses it,
 note this gap and suggest a follow-up issue for the remaining work.
 

package/{src → .github}/agents/agent-supervisor.md

@@ -85,6 +85,15 @@ Check the Recent Activity log and Recently Closed Issues for patterns:
 **Dedup deadlock recovery:**
 - If your issue creation is blocked by the dedup guard (similar to a recently closed issue), do NOT retry the same issue title. Instead, create an issue with a different scope — e.g. "fix: resolve failing tests on main" or "fix: align code and test expectations" — with both the `instability` and `ready` labels. This sidesteps the dedup guard and directs the dev job to the actual problem. The `instability` label gives the issue mechanical priority over other `ready` issues.
 
+## Trend Analysis (intentïon.md)
+
+When intentïon.md is available (attached), examine it for strategic insights:
+
+1. **Iteration trends** — look for patterns in recent transforms: are they productive (code landed, tests passing) or spinning (same issues re-opened, budget consumed with no progress)? Adjust your dispatch strategy accordingly.
+2. **Recurring failures** — if the same test or feature keeps failing across iterations, don't just re-dispatch transform. Create a targeted issue that explicitly addresses the root cause, or dispatch fix-code instead.
+3. **Budget trajectory** — correlate the narrative with the budget shown in context. If budget is being consumed rapidly with little progress, switch to conservative actions (review, maintain) rather than speculative transforms.
+4. **Mission progress** — use the narrative to assess how close the mission is to completion. If most acceptance criteria are met, focus on closing gaps rather than broad transforms.
+
 ## Discussions Awareness
 
 Check the Recent Activity log for discussion bot referrals (lines containing `discussion-request-supervisor`). These indicate a user asked the bot something that requires supervisor action. **Prioritise responding to these referrals.**

package/.github/workflows/agentic-lib-test.yml

@@ -130,12 +130,14 @@ jobs:
           path: SCREENSHOT_INDEX.png
           if-no-files-found: ignore
 
-      - name: Push screenshot to agentic-lib-logs branch
+      - name: Push screenshot to log branch
         if: |
           github.ref == 'refs/heads/main' &&
           github.repository != 'xn-intenton-z2a/agentic-lib' &&
           (github.event_name == 'schedule' || inputs.push-screenshot == 'true' || inputs.push-screenshot == true)
-        run: bash .github/agentic-lib/scripts/push-to-logs.sh SCREENSHOT_INDEX.png
+        env:
+          LOG_BRANCH: ${{ inputs.log-branch || 'agentic-lib-logs' }}
+        run: bash .github/agentic-lib/scripts/push-to-logs.sh "${SCREENSHOT_FILE:-SCREENSHOT_INDEX.png}"
 
   # ─── Dispatch fix workflow when unit tests fail on main ─────────────
   dispatch-fix:

package/.github/workflows/agentic-lib-workflow.yml

@@ -148,13 +148,13 @@ jobs:
         run: |
           MODEL='${{ inputs.model }}'
           if [ -z "$MODEL" ] && [ -f "${{ env.configPath }}" ]; then
-            TOML_MODEL=$(grep '^\s*model' "${{ env.configPath }}" | head -1 | sed 's/.*= *"\([^"]*\)".*/\1/')
+            TOML_MODEL=$(grep '^\s*model' "${{ env.configPath }}" | head -1 | sed 's/.*= *"\([^"]*\)".*/\1/' || true)
             MODEL="${TOML_MODEL}"
           fi
           echo "model=${MODEL:-gpt-5-mini}" >> $GITHUB_OUTPUT
           PROFILE='${{ inputs.profile }}'
           if [ -z "$PROFILE" ] && [ -f "${{ env.configPath }}" ]; then
-            TOML_PROFILE=$(grep '^\s*profile' "${{ env.configPath }}" | head -1 | sed 's/.*= *"\([^"]*\)".*/\1/')
+            TOML_PROFILE=$(grep '^\s*profile' "${{ env.configPath }}" | head -1 | sed 's/.*= *"\([^"]*\)".*/\1/' || true)
             PROFILE="${TOML_PROFILE}"
           fi
           echo "profile=${PROFILE}" >> $GITHUB_OUTPUT
@@ -178,6 +178,18 @@ jobs:
           echo "dry-run=${DRY_RUN}" >> $GITHUB_OUTPUT
           CONFIG='${{ inputs.config-path }}'
           echo "config-path=${CONFIG:-${{ env.configPath }}}" >> $GITHUB_OUTPUT
+          # Bot config: log file, log branch, screenshot file
+          LOG_FILE=""
+          LOG_BRANCH=""
+          SCREENSHOT=""
+          if [ -f "${{ env.configPath }}" ]; then
+            LOG_FILE=$(grep '^\s*log-file' "${{ env.configPath }}" | head -1 | sed 's/.*= *"\([^"]*\)".*/\1/' || true)
+            LOG_BRANCH=$(grep '^\s*log-branch' "${{ env.configPath }}" | head -1 | sed 's/.*= *"\([^"]*\)".*/\1/' || true)
+            SCREENSHOT=$(grep '^\s*screenshot-file' "${{ env.configPath }}" | head -1 | sed 's/.*= *"\([^"]*\)".*/\1/' || true)
+          fi
+          echo "log-file=${LOG_FILE:-intentïon.md}" >> $GITHUB_OUTPUT
+          echo "log-branch=${LOG_BRANCH:-agentic-lib-logs}" >> $GITHUB_OUTPUT
+          echo "screenshot-file=${SCREENSHOT:-SCREENSHOT_INDEX.png}" >> $GITHUB_OUTPUT
     outputs:
       model: ${{ steps.normalise.outputs.model }}
       profile: ${{ steps.normalise.outputs.profile }}
@@ -188,6 +200,9 @@ jobs:
       pr-number: ${{ steps.normalise.outputs.pr-number }}
       dry-run: ${{ steps.normalise.outputs.dry-run }}
       config-path: ${{ steps.normalise.outputs.config-path }}
+      log-file: ${{ steps.normalise.outputs.log-file }}
+      log-branch: ${{ steps.normalise.outputs.log-branch }}
+      screenshot-file: ${{ steps.normalise.outputs.screenshot-file }}
 
   # ─── PR Cleanup: merge/close/delete stale PRs and branches ─────────
   pr-cleanup:
@@ -357,6 +372,9 @@ jobs:
       - name: Gather telemetry
         id: gather
         uses: actions/github-script@v8
+        env:
+          LOG_FILE: ${{ needs.params.outputs.log-file }}
+          LOG_BRANCH: ${{ needs.params.outputs.log-branch }}
         with:
           script: |
             const fs = require('fs');
@@ -460,23 +478,25 @@ jobs:
             const missionComplete = fs.existsSync('MISSION_COMPLETE.md');
             const missionFailed = fs.existsSync('MISSION_FAILED.md');
 
-            // Activity log stats (fetched from agentic-lib-logs branch)
+            // Activity log stats (fetched from log branch)
+            const logFile = process.env.LOG_FILE || 'intentïon.md';
+            const logBranch = process.env.LOG_BRANCH || 'agentic-lib-logs';
             let activityStats = null;
             try {
-              // Try agentic-lib-logs branch first (new location), fall back to local file (legacy)
+              // Try log branch first, fall back to local file (legacy)
               let log = null;
-              for (const logFile of ['intenti\u00F6n.md', 'intention.md']) {
+              for (const lf of [logFile, 'intention.md']) {
                 try {
                   const { data } = await github.rest.repos.getContent({
-                    owner, repo, path: logFile, ref: 'agentic-lib-logs',
+                    owner, repo, path: lf, ref: logBranch,
                   });
                   log = Buffer.from(data.content, 'base64').toString('utf8');
                   break;
                 } catch { /* try next */ }
               }
-              // Fall back to local file (for repos not yet using agentic-lib-logs branch)
+              // Fall back to local file
               if (!log) {
-                const logPath = fs.existsSync('intenti\u00F6n.md') ? 'intenti\u00F6n.md' : (fs.existsSync('intention.md') ? 'intention.md' : null);
+                const logPath = fs.existsSync(logFile) ? logFile : (fs.existsSync('intention.md') ? 'intention.md' : null);
                 if (logPath) log = fs.readFileSync(logPath, 'utf8');
               }
               if (log) {
@@ -570,10 +590,14 @@ jobs:
           fetch-depth: 0
           ref: ${{ inputs.ref || github.sha }}
 
-      - name: Fetch log from agentic-lib-logs branch
+      - name: Fetch log and screenshot from log branch
+        env:
+          LOG_FILE: ${{ needs.params.outputs.log-file }}
+          LOG_BRANCH: ${{ needs.params.outputs.log-branch }}
+          SCREENSHOT_FILE: ${{ needs.params.outputs.screenshot-file }}
         run: |
-          for f in "intentïon.md" "intention.md"; do
-            git show "origin/agentic-lib-logs:${f}" > "$f" 2>/dev/null || true
+          for f in "${LOG_FILE}" "intention.md" "${SCREENSHOT_FILE}"; do
+            git show "origin/${LOG_BRANCH}:${f}" > "$f" 2>/dev/null || true
           done
 
       - name: Check mission-complete signal
@@ -680,9 +704,11 @@ jobs:
         id: get-sha
         run: echo "sha=$(git rev-parse HEAD)" >> $GITHUB_OUTPUT
 
-      - name: Push log to agentic-lib-logs branch
+      - name: Push log to log branch
         if: github.repository != 'xn-intenton-z2a/agentic-lib' && needs.params.outputs.dry-run != 'true'
-        run: bash .github/agentic-lib/scripts/push-to-logs.sh "intentïon.md" "intention.md"
+        env:
+          LOG_BRANCH: ${{ needs.params.outputs.log-branch }}
+        run: bash .github/agentic-lib/scripts/push-to-logs.sh "${{ needs.params.outputs.log-file }}" "intention.md"
 
   # ─── Director: LLM evaluates mission status (complete/failed/in-progress) ──
   director:
@@ -701,10 +727,14 @@ jobs:
           fetch-depth: 0
           ref: ${{ needs.maintain.outputs.commit-sha || github.sha }}
 
-      - name: Fetch log from agentic-lib-logs branch
+      - name: Fetch log and screenshot from log branch
+        env:
+          LOG_FILE: ${{ needs.params.outputs.log-file }}
+          LOG_BRANCH: ${{ needs.params.outputs.log-branch }}
+          SCREENSHOT_FILE: ${{ needs.params.outputs.screenshot-file }}
         run: |
-          for f in "intentïon.md" "intention.md"; do
-            git show "origin/agentic-lib-logs:${f}" > "$f" 2>/dev/null || true
+          for f in "${LOG_FILE}" "intention.md" "${SCREENSHOT_FILE}"; do
+            git show "origin/${LOG_BRANCH}:${f}" > "$f" 2>/dev/null || true
           done
 
       - uses: actions/setup-node@v6
@@ -736,9 +766,11 @@ jobs:
           instructions: ".github/agents/agent-director.md"
           model: ${{ needs.params.outputs.model }}
 
-      - name: Push log to agentic-lib-logs branch
+      - name: Push log to log branch
         if: github.repository != 'xn-intenton-z2a/agentic-lib' && needs.params.outputs.dry-run != 'true'
-        run: bash .github/agentic-lib/scripts/push-to-logs.sh "intentïon.md" "intention.md"
+        env:
+          LOG_BRANCH: ${{ needs.params.outputs.log-branch }}
+        run: bash .github/agentic-lib/scripts/push-to-logs.sh "${{ needs.params.outputs.log-file }}" "intention.md"
 
   # ─── Supervisor: LLM decides what to do (after director evaluates) ──
   supervisor:
@@ -756,10 +788,14 @@ jobs:
           fetch-depth: 0
           ref: ${{ needs.maintain.outputs.commit-sha || github.sha }}
 
-      - name: Fetch log from agentic-lib-logs branch
+      - name: Fetch log and screenshot from log branch
+        env:
+          LOG_FILE: ${{ needs.params.outputs.log-file }}
+          LOG_BRANCH: ${{ needs.params.outputs.log-branch }}
+          SCREENSHOT_FILE: ${{ needs.params.outputs.screenshot-file }}
         run: |
-          for f in "intentïon.md" "intention.md"; do
-            git show "origin/agentic-lib-logs:${f}" > "$f" 2>/dev/null || true
+          for f in "${LOG_FILE}" "intention.md" "${SCREENSHOT_FILE}"; do
+            git show "origin/${LOG_BRANCH}:${f}" > "$f" 2>/dev/null || true
           done
 
       - uses: actions/setup-node@v6
@@ -791,9 +827,11 @@ jobs:
           instructions: ".github/agents/agent-supervisor.md"
           model: ${{ needs.params.outputs.model }}
 
-      - name: Push log to agentic-lib-logs branch
+      - name: Push log to log branch
         if: github.repository != 'xn-intenton-z2a/agentic-lib' && needs.params.outputs.dry-run != 'true'
-        run: bash .github/agentic-lib/scripts/push-to-logs.sh "intentïon.md" "intention.md"
+        env:
+          LOG_BRANCH: ${{ needs.params.outputs.log-branch }}
+        run: bash .github/agentic-lib/scripts/push-to-logs.sh "${{ needs.params.outputs.log-file }}" "intention.md"
 
   # ─── Fix stuck PRs with failing checks ─────────────────────────────
   fix-stuck:
@@ -1187,10 +1225,14 @@ jobs:
           token: ${{ secrets.GITHUB_TOKEN }}
           ref: ${{ needs.maintain.outputs.commit-sha || github.sha }}
 
-      - name: Fetch log from agentic-lib-logs branch
+      - name: Fetch log and screenshot from log branch
+        env:
+          LOG_FILE: ${{ needs.params.outputs.log-file }}
+          LOG_BRANCH: ${{ needs.params.outputs.log-branch }}
+          SCREENSHOT_FILE: ${{ needs.params.outputs.screenshot-file }}
         run: |
-          for f in "intentïon.md" "intention.md"; do
-            git show "origin/agentic-lib-logs:${f}" > "$f" 2>/dev/null || true
+          for f in "${LOG_FILE}" "intention.md" "${SCREENSHOT_FILE}"; do
+            git show "origin/${LOG_BRANCH}:${f}" > "$f" 2>/dev/null || true
           done
 
       - uses: actions/setup-node@v6
@@ -1357,9 +1399,11 @@ jobs:
           commit-message: "agentic-step: transform issue #${{ steps.issue.outputs.issue-number }}"
           push-ref: ${{ steps.branch.outputs.branchName }}
 
-      - name: Push log to agentic-lib-logs branch
+      - name: Push log to log branch
         if: github.repository != 'xn-intenton-z2a/agentic-lib' && needs.params.outputs.dry-run != 'true'
-        run: bash .github/agentic-lib/scripts/push-to-logs.sh "intentïon.md" "intention.md"
+        env:
+          LOG_BRANCH: ${{ needs.params.outputs.log-branch }}
+        run: bash .github/agentic-lib/scripts/push-to-logs.sh "${{ needs.params.outputs.log-file }}" "intention.md"
 
       - name: Create PR and attempt merge
         if: github.repository != 'xn-intenton-z2a/agentic-lib' && steps.issue.outputs.issue-number != '' && needs.params.outputs.dry-run != 'true' && steps.pre-commit-test.outputs.tests-passed == 'true' && steps.pre-commit-behaviour-test.outputs.tests-passed != 'false'

package/README.md

@@ -195,16 +195,14 @@ log-file = "intentïon.md"
 
 The `profile` setting controls all tuning defaults. Three profiles are built in:
 
-| Profile | Budget | Source scan | Issues | Best for |
-|---------|--------|-------------|--------|----------|
-| `min` | 4 cycles | 3 files, 1000 chars | 5, 14d stale | CI testing, quick validation |
-| `recommended` | 8 cycles | 10 files, 5000 chars | 20, 30d stale | Balanced cost/quality |
-| `max` | 32 cycles | 50 files, 20000 chars | 100, 90d stale | Complex missions |
+| Profile | Budget | Issues | Best for |
+|---------|--------|--------|----------|
+| `min` | 16 cycles | 5, 14d stale | CI testing, quick validation |
+| `recommended` | 32 cycles | 20, 30d stale | Balanced cost/quality |
+| `max` | 128 cycles | 100, 90d stale | Complex missions |
 
 Override individual knobs in `[tuning]` to deviate from a profile. Limits (`[limits]`) also scale with the profile.
 
-The legacy YAML config at `.github/agentic-lib/agents/agentic-lib.yml` is also supported as a fallback.
-
 ## The `agentic-step` Action
 
 The core of the system is a single GitHub Action that handles all autonomous tasks:

package/agentic-lib.toml

@@ -52,14 +52,8 @@ model = "gpt-5-mini"
 # reasoning-effort = "low"       # low | medium | high | none (none = disable entirely)
 infinite-sessions = false        # set to true for long sessions with compaction
 # transformation-budget = 4      # max code-changing cycles per run (0 = unlimited)
-# max-feature-files = 3
-# max-source-files = 3
-# max-source-chars = 1000
-# max-test-chars = 500
 # max-issues = 5
-# issue-body-limit = 200
 # stale-days = 14
-# max-summary-chars = 500
 # max-discussion-comments = 5
 
 # ─── Profile Definitions ────────────────────────────────────────────
@@ -68,19 +62,13 @@ infinite-sessions = false        # set to true for long sessions with compaction
 # [tuning] or [limits] to deviate from the active profile.
 
 [profiles.min]
-# ~10% of gpt-4.1 context (128K). Fast, cheap — CI testing, quick validation.
+# Fast, cheap — CI testing, quick validation.
 reasoning-effort = "low"                # low | medium | high | none
 infinite-sessions = false               # enable session compaction for long runs
 transformation-budget = 16             # max code-changing cycles per run
-max-feature-files = 3                      # max feature files included in prompts
-max-source-files = 3                        # max source files included in prompts
-max-source-chars = 4000                  # max chars of each source file's content included in prompts
-max-test-chars = 2000                     # max chars of each test file's content included in prompts
-max-issues = 5                        # max open issues included in prompts
-issue-body-limit = 1000                 # max chars of each issue's body text included in prompts
-stale-days = 14                        # days since last activity before an issue is considered stale and excluded
-max-summary-chars = 2000                 # max chars of each document's summary included in prompts
-max-discussion-comments = 3               # max recent discussion comments (newest first) included in prompts
+max-issues = 5                        # max open issues fetched from GitHub
+stale-days = 14                        # days since last activity before an issue is considered stale
+max-discussion-comments = 3               # max recent discussion comments fetched
 max-feature-issues = 1                     # max concurrent feature development issues
 max-maintenance-issues = 1                 # max concurrent maintenance issues
 max-attempts-per-branch = 2               # max transform attempts before abandoning a branch
@@ -89,19 +77,13 @@ features-limit = 2                     # max feature files in features/ director
 library-limit = 8                      # max library entries in library/ directory
 
 [profiles.recommended]
-# ~30% of claude-sonnet-4 context (216K). Balanced — good results, default for consumer repos.
+# Balanced — good results, default for consumer repos.
 reasoning-effort = "medium"            # low | medium | high | none
 infinite-sessions = true               # enable session compaction for long runs
 transformation-budget = 32             # max code-changing cycles per run
-max-feature-files = 10                     # max feature files included in prompts
-max-source-files = 10                       # max source files included in prompts
-max-source-chars = 8000                  # max chars of each source file's content included in prompts
-max-test-chars = 5000                    # max chars of each test file's content included in prompts
-max-issues = 15                       # max open issues included in prompts
-issue-body-limit = 2000                 # max chars of each issue's body text included in prompts
-stale-days = 30                        # days since last activity before an issue is considered stale and excluded
-max-summary-chars = 5000                # max chars of each document's summary included in prompts
-max-discussion-comments = 10              # max recent discussion comments (newest first) included in prompts
+max-issues = 15                       # max open issues fetched from GitHub
+stale-days = 30                        # days since last activity before an issue is considered stale
+max-discussion-comments = 10              # max recent discussion comments fetched
 max-feature-issues = 2                     # max concurrent feature development issues
 max-maintenance-issues = 1                 # max concurrent maintenance issues
 max-attempts-per-branch = 3               # max transform attempts before abandoning a branch
@@ -110,19 +92,13 @@ features-limit = 4                     # max feature files in features/ director
 library-limit = 32                     # max library entries in library/ directory
 
 [profiles.max]
-# ~90% of gpt-5-mini context (264K). Thorough — maximum context for complex missions.
+# Thorough — maximum context for complex missions.
 reasoning-effort = "high"              # low | medium | high | none
 infinite-sessions = true               # enable session compaction for long runs
 transformation-budget = 128           # max code-changing cycles per run
-max-feature-files = 20                     # max feature files included in prompts
-max-source-files = 25                       # max source files included in prompts
-max-source-chars = 12000                 # max chars of each source file's content included in prompts
-max-test-chars = 8000                   # max chars of each test file's content included in prompts
-max-issues = 30                      # max open issues included in prompts
-issue-body-limit = 3000                # max chars of each issue's body text included in prompts
-stale-days = 90                        # days since last activity before an issue is considered stale and excluded
-max-summary-chars = 8000               # max chars of each document's summary included in prompts
-max-discussion-comments = 20              # max recent discussion comments (newest first) included in prompts
+max-issues = 30                      # max open issues fetched from GitHub
+stale-days = 90                        # days since last activity before an issue is considered stale
+max-discussion-comments = 20              # max recent discussion comments fetched
 max-feature-issues = 4                     # max concurrent feature development issues
 max-maintenance-issues = 2                 # max concurrent maintenance issues
 max-attempts-per-branch = 5               # max transform attempts before abandoning a branch
@@ -133,9 +109,11 @@ library-limit = 64                     # max library entries in library/ directo
 [mission-complete]
 # Thresholds for deterministic mission-complete declaration.
 # All conditions must be met simultaneously.
-min-resolved-issues = 3            # minimum closed-as-RESOLVED issues since init
-min-dedicated-tests = 1            # minimum test files that import from src/lib/
+min-resolved-issues = 2            # minimum closed-as-RESOLVED issues since init
+min-dedicated-tests = 0            # minimum test files that import from src/lib/
 max-source-todos = 0               # max TODO comments allowed in ./src (0 = none)
 
 [bot]
 log-file = "test/intentïon.md"                  #@dist "intentïon.md"
+log-branch = "agentic-lib-logs"
+screenshot-file = "SCREENSHOT_INDEX.png"