ecological-agent-skills 3.1.0

package/AGENT_CONTEXT.md

@@ -0,0 +1,191 @@
+# AGENT_CONTEXT.md
+
+This file is written for AI agents, not humans.
+Read this file first before performing any task in this repository.
+Do not skip sections. Every rule here is enforced at runtime.
+
+---
+
+## 1. How to Use This Repository
+
+This repository is a modular skill library for quantitative ecology.
+It is structured for use by AI agents operating within agent frameworks
+(Claude Code, Gemini CLI, GitHub Copilot, Cursor, or any compatible system).
+
+**Skills** are self-contained analysis modules. Each skill:
+- Lives at `skills/<skill-id>/`
+- Is described by `skills/<skill-id>/SKILL.md`
+- Contains scripts in `scripts/`, reference material in `resources/`, and
+  worked examples in `examples/`
+
+**Workflows** are ordered sequences of skills for complete project pipelines.
+Each workflow lives at `workflows/<workflow-id>/WORKFLOW.md`.
+
+**Rules — read before invoking any skill:**
+
+1. Read `skills/SKILL_INDEX.json` first to identify the correct skill for the task.
+   Do not rely on skill names alone; use `trigger_keywords` and `domain` fields.
+2. Read the full `SKILL.md` of the selected skill before executing any step.
+   Do not skip to the scripts.
+3. Check `min_inputs` in the skill index. If a required input is missing, request it
+   from the user before proceeding.
+4. After completing each analytical step, write decisions to `decision_log.md` as
+   required by the `reproducible-ecology-pipeline` skill.
+5. Validate outputs listed in `primary_outputs` before marking a skill as complete.
+
+---
+
+## 2. Canonical Invocation Order by Project Type
+
+| Project Type | Step 1 | Step 2 | Step 3 | Step 4 | Step 5 |
+|---|---|---|---|---|---|
+| **SDM — complete** | ecological-data-foundation | geoprocessing-for-ecology | predictive-modeling-best-practices | species-distribution-modeling | model-validation-and-uncertainty |
+| **SDM — projection only** | geoprocessing-for-ecology | species-distribution-modeling | model-validation-and-uncertainty | — | — |
+| **Ecological impact (BACI)** | ecological-data-foundation | ecological-impact-assessment | biostatistics-workbench | model-validation-and-uncertainty | reproducible-ecology-pipeline |
+| **Community analysis** | ecological-data-foundation | community-ecology-ordination | biostatistics-workbench | reproducible-ecology-pipeline | — |
+| **Occupancy and detection** | ecological-data-foundation | occupancy-and-detection | biostatistics-workbench | reproducible-ecology-pipeline | — |
+| **Environmental time series** | ecological-data-foundation | environmental-time-series | biostatistics-workbench | reproducible-ecology-pipeline | — |
+| **Ecosystem services** | ecological-data-foundation | geoprocessing-for-ecology | ecosystem-services-assessment | model-validation-and-uncertainty | reproducible-ecology-pipeline |
+| **Multispecies screening** | ecological-data-foundation | geoprocessing-for-ecology | species-distribution-modeling (loop) | model-validation-and-uncertainty | reproducible-ecology-pipeline |
+
+For multispecies projects always read `workflows/run-multispecies-screening/WORKFLOW.md`
+before starting. It contains the priority classification logic.
+
+---
+
+## 3. Disambiguation Rules for Overlapping Skills
+
+Apply exactly one rule per ambiguous situation. Do not invoke two skills for the
+same analytical question.
+
+### biostatistics-workbench vs community-ecology-ordination
+- Use **biostatistics-workbench** when the question involves a single response variable
+  (abundance, richness, biomass) and one or more predictors.
+- Use **community-ecology-ordination** when the question involves species composition
+  across multiple sites (a species × site matrix is the primary input).
+
+### ecological-data-foundation vs geoprocessing-for-ecology
+- Use **ecological-data-foundation** for cleaning, deduplicating, validating, and
+  standardising occurrence records or survey tables.
+- Use **geoprocessing-for-ecology** for spatial operations on rasters and vectors:
+  reprojection, masking, stacking, buffer creation, spatial extraction.
+- If the task involves occurrence records AND spatial extraction, invoke
+  ecological-data-foundation first, then geoprocessing-for-ecology.
+
+### model-validation-and-uncertainty vs predictive-modeling-best-practices
+- Use **predictive-modeling-best-practices** BEFORE fitting any model:
+  predictor selection, collinearity reduction, CV strategy design, bias correction.
+- Use **model-validation-and-uncertainty** AFTER fitting any model:
+  performance metrics, calibration, uncertainty quantification, extrapolation risk.
+- Never reverse this order.
+
+### ecological-impact-assessment vs biostatistics-workbench
+- Use **ecological-impact-assessment** when the data has explicit temporal structure
+  (before/after) and spatial structure (control/impact sites), or when computing
+  landscape fragmentation or pressure indices.
+- Use **biostatistics-workbench** for statistical comparisons that do not involve
+  a BACI design or landscape metrics.
+
+### species-distribution-modeling vs occupancy-and-detection
+- Use **species-distribution-modeling** when the goal is a spatial suitability surface
+  using presence-only or presence-absence records and environmental predictors.
+- Use **occupancy-and-detection** when the goal is to estimate occupancy probability (ψ)
+  and detection probability (p) using replicated-visit detection history data.
+
+### ecological-data-foundation vs reproducible-ecology-pipeline
+- Use **ecological-data-foundation** for data cleaning and QA at the start of the pipeline.
+- Use **reproducible-ecology-pipeline** for logging decisions, generating file manifests,
+  and ensuring auditability — invoke at the end of each major analytical phase.
+
+---
+
+## 4. When NOT to Invoke Any Skill
+
+Do not invoke any skill for the following situations. Answer directly.
+
+- **Conceptual questions about methods**: "What is MaxEnt?", "How does NMDS work?"
+  → Explain directly. No skill invocation.
+- **Interpretation of already-computed results**: "What does an AUC of 0.75 mean for
+  my model?" → Interpret directly. Do not re-run scripts.
+- **Code review of existing scripts**: "Is this R script correct?" → Review and respond
+  directly. Do not re-execute the script.
+- **n < 10 records for any modelling task**: Do not invoke any modelling skill.
+  Communicate the limitation to the user explicitly. Suggest options (literature review,
+  expert knowledge, qualitative assessment).
+
+---
+
+## 5. Minimum Sample Size Requirements
+
+If the available n is below the minimum absolute threshold, do not proceed with modelling.
+State the limitation, explain why modelling is unreliable, and suggest alternatives.
+
+| Analysis Type | n minimum (absolute) | n recommended | Action if below minimum |
+|---|---|---|---|
+| SDM (MaxEnt/ensemble) | 10 occurrences | ≥ 30 | Do not fit model; suggest literature-based range map |
+| Occupancy model (single-season) | 15 sites | ≥ 30 sites, ≥ 3 visits | Do not fit model; use naive occupancy with explicit caveat |
+| BACI (mixed model) | 5 control + 5 impact sites, ≥ 3 time periods | ≥ 10 per group | Report power deficit; fit model with explicit uncertainty caveat |
+| Community ordination (NMDS/PERMANOVA) | 5 sites | ≥ 10 sites | Reduce dimensionality; note low power |
+| GLM (single predictor) | 20 observations | ≥ 50 | Fit with caution; report wide CIs explicitly |
+| PVA (population viability) | 5 years of count data | ≥ 10 years | Do not run stochastic PVA; use deterministic model only |
+
+---
+
+## 6. Scaling Between Simple and Complex Projects
+
+Apply the appropriate routing based on project complexity:
+
+**Simple project** (1 species, 1 analysis, no pipeline):
+→ Invoke the single relevant skill directly.
+→ Example: "Clean this occurrence CSV" → ecological-data-foundation only.
+
+**Medium project** (1 species, full analytical pipeline):
+→ Follow the canonical invocation order from Section 2.
+→ Read the corresponding workflow WORKFLOW.md before starting.
+→ Example: Complete SDM → read `workflows/run-sdm-study/WORKFLOW.md`.
+
+**Complex project** (multiple species, multiple analyses, or iterative pipeline):
+→ Read `workflows/run-multispecies-screening/WORKFLOW.md` first.
+→ Use `future::plan(multisession)` in R for species-level parallelisation.
+→ Prioritise species using the High/Medium/Low classification defined in that workflow.
+→ Do not proceed to projection for Medium or Low priority species until High priority
+  species have passed model validation.
+
+---
+
+## 7. File Conventions of This Repository
+
+### Where to put input data
+- Place raw input files in `data/raw/` (create if absent).
+- Place cleaned, analysis-ready files in `data/processed/`.
+- GBIF downloads go in `data/raw/gbif/`.
+
+### Where to write outputs
+- Model outputs: `outputs/<skill-id>/` (create per skill, per species if needed).
+- Reports: `outputs/reports/`.
+- Figures: `outputs/figures/`.
+- Logs: `outputs/logs/`.
+
+### Naming conventions
+- Occurrence files: `occurrences_<species-slug>_<source>_<YYYYMMDD>.csv`
+- Raster outputs: `<variable>_<species-slug>_<scenario>_<year>.tif`
+- Reports: `<skill-id>_report_<YYYYMMDD>.md`
+- Follow `snake_case` for all file names. No spaces.
+
+### Recording decisions
+After each analytical step, append to `decision_log.md` at the project root:
+```
+## [YYYY-MM-DD] <Skill ID> — <brief description>
+- Decision: <what was decided>
+- Rationale: <why>
+- Alternatives considered: <what else was evaluated>
+- Output files: <list of files generated>
+```
+If `decision_log.md` does not exist, create it before appending.
+
+### Checking the skill index
+Before invoking any skill, execute this lookup:
+1. Open `skills/SKILL_INDEX.json`.
+2. Search `trigger_keywords` for terms matching the user request.
+3. Check `depends_on_skills` — all dependencies must have been completed first.
+4. Check `decision_points` — verify n thresholds and quality gates before proceeding.

package/CATALOG.md

@@ -0,0 +1,329 @@
+# Skill Catalog — ecological-agent-skills
+
+A quick-reference index for all 17 skills. Each row summarises the skill's domain, when to invoke it, expected inputs, expected outputs, and which workflows depend on it.
+
+---
+
+## Index
+
+| # | Skill | Domain | Phase |
+|---|-------|--------|-------|
+| 1 | [ecological-data-foundation](#1-ecological-data-foundation) | Data management | 1 |
+| 2 | [geoprocessing-for-ecology](#2-geoprocessing-for-ecology) | Spatial analysis | 1 |
+| 3 | [biostatistics-workbench](#3-biostatistics-workbench) | Statistics | 1 |
+| 4 | [predictive-modeling-best-practices](#4-predictive-modeling-best-practices) | ML / modeling | 1 |
+| 5 | [model-validation-and-uncertainty](#5-model-validation-and-uncertainty) | Evaluation | 2 |
+| 6 | [species-distribution-modeling](#6-species-distribution-modeling) | SDM / ENM | 2 |
+| 7 | [occupancy-and-detection](#7-occupancy-and-detection) | Occupancy models | 3 |
+| 8 | [community-ecology-ordination](#8-community-ecology-ordination) | Community ecology | 3 |
+| 9 | [ecological-impact-assessment](#9-ecological-impact-assessment) | Impact & pressure | 2 |
+| 10 | [environmental-time-series](#10-environmental-time-series) | Time series | 2 |
+| 11 | [ecosystem-services-assessment](#11-ecosystem-services-assessment) | ES valuation | 3 |
+| 12 | [reproducible-ecology-pipeline](#12-reproducible-ecology-pipeline) | Reproducibility | 1 |
+| 13 | [camera-trap-processing](#13-camera-trap-processing) | Camera traps / wildlife monitoring | 4 |
+| 14 | [acoustic-monitoring](#14-acoustic-monitoring) | Bioacoustics / soundscape ecology | 4 |
+| 15 | [landscape-connectivity](#15-landscape-connectivity) | Landscape connectivity / corridors | 4 |
+| 16 | [population-viability-analysis](#16-population-viability-analysis) | Population viability / extinction risk | 4 |
+| 17 | [spatial-prioritization](#17-spatial-prioritization) | Conservation planning / reserve design | 4 |
+
+---
+
+## Skill Details
+
+### 1. ecological-data-foundation
+**Domain:** Data ingestion, cleaning, schema validation, metadata, initial QA  
+**When to use:** At the start of any quantitative ecology project, before spatial operations or modeling.  
+**Inputs:** Raw occurrence records, field survey tables, environmental rasters, sensor exports, CSV/XLSX/GDB files  
+**Outputs:** Cleaned, validated dataset; QA report; metadata record; standardised schema  
+**Used by workflows:** all  
+
+---
+
+### 2. geoprocessing-for-ecology
+**Domain:** Coordinate reference systems, raster/vector operations, spatial extraction  
+**When to use:** Whenever spatial data needs to be reprojected, clipped, buffered, intersected, or used to extract environmental values.  
+**Inputs:** Shapefiles, GeoTIFFs, occurrence points, study area polygon  
+**Outputs:** Reprojected/masked layers, extracted raster values, spatial join tables, buffer polygons  
+**Used by workflows:** run-sdm-study, assess-ecological-impact, build-fire-risk-map, analyze-environmental-change, assess-ecosystem-services  
+
+---
+
+### 3. biostatistics-workbench
+**Domain:** Hypothesis testing, GLM/GLMM, assumption checking, effect size, confidence intervals  
+**When to use:** For inferential statistics, model selection between statistical models, or validating assumptions of any test.  
+**Inputs:** Tabular ecological data, response and predictor variables  
+**Outputs:** Test statistics, p-values, effect sizes, model comparison tables, assumption diagnostics  
+**Used by workflows:** assess-ecological-impact, analyze-community-structure, run-occupancy-analysis, assess-ecosystem-services  
+
+---
+
+### 4. predictive-modeling-best-practices
+**Domain:** Train/test split, cross-validation, hyperparameter tuning, collinearity, overfitting, data leakage  
+**When to use:** Before fitting any machine learning or algorithmic model (SDM, random forest, BRT, etc.).  
+**Inputs:** Feature matrix, target variable, candidate model list  
+**Outputs:** CV strategy, tuning results, collinearity report, leakage audit  
+**Used by workflows:** run-sdm-study, build-fire-risk-map  
+
+---
+
+### 5. model-validation-and-uncertainty
+**Domain:** Performance metrics, calibration, sensitivity analysis, external validation, uncertainty quantification  
+**When to use:** After any model is fitted; always required before reporting results.  
+**Inputs:** Model object, validation dataset, predictions  
+**Outputs:** AUC/TSS/RMSE/etc., calibration plots, sensitivity report, uncertainty maps/intervals  
+**Used by workflows:** run-sdm-study, assess-ecological-impact, analyze-community-structure, build-fire-risk-map, run-occupancy-analysis  
+
+---
+
+### 6. species-distribution-modeling
+**Domain:** SDM/ENM full pipeline (MaxEnt, BRT, ensemble, projection)  
+**When to use:** To model the potential or realised distribution of one or more species.  
+**Inputs:** Cleaned occurrence records, environmental predictor stack, study area extent  
+**Outputs:** Suitability map, variable importance, model ensemble, projection under scenarios  
+**Used by workflows:** run-sdm-study  
+
+---
+
+### 7. occupancy-and-detection
+**Domain:** Single-season and dynamic occupancy models, imperfect detection, replicate surveys  
+**When to use:** When detection is imperfect and the goal is to estimate occupancy probability separately from detection probability.  
+**Inputs:** Detection/non-detection matrix, site covariates, observation covariates  
+**Outputs:** Occupancy estimates (ψ), detection estimates (p), covariate effects, AIC table  
+**Used by workflows:** run-occupancy-analysis  
+
+---
+
+### 8. community-ecology-ordination
+**Domain:** NMDS, PCA, PCoA, diversity indices, cluster analysis, community composition  
+**When to use:** For multivariate analysis of species assemblages, beta diversity, or community structure comparisons.  
+**Inputs:** Species × site matrix, environmental metadata  
+**Outputs:** Ordination plots, stress/eigenvalues, diversity metrics, cluster dendrograms, PERMANOVA results  
+**Used by workflows:** analyze-community-structure  
+
+---
+
+### 9. ecological-impact-assessment
+**Domain:** BACI design, before/after analysis, landscape fragmentation, anthropogenic pressure synthesis  
+**When to use:** To quantify the effect of a disturbance, land-use change, or management intervention on ecological indicators.  
+**Inputs:** Pre/post datasets, control/impact site data, pressure layers  
+**Outputs:** Impact magnitude estimates, fragmentation metrics, pressure index, synthesis narrative  
+**Used by workflows:** assess-ecological-impact, build-fire-risk-map, analyze-environmental-change  
+
+---
+
+### 10. environmental-time-series
+**Domain:** Trend detection, seasonality decomposition, breakpoint analysis, anomaly detection, recovery trajectories  
+**When to use:** For any temporal ecological signal: NDVI, temperature, rainfall, species abundance index, etc.  
+**Inputs:** Time-indexed environmental or ecological variable (tabular or raster stack)  
+**Outputs:** Trend estimate, seasonal component, breakpoint dates, anomaly flags, recovery rate  
+**Used by workflows:** build-fire-risk-map, analyze-environmental-change  
+
+---
+
+### 11. ecosystem-services-assessment
+**Domain:** ES indicators, provisioning/regulating/cultural service quantification, trade-off analysis  
+**When to use:** To evaluate and spatially represent ecosystem services across a landscape.  
+**Inputs:** Land cover map, biophysical data, socioeconomic layers, study area  
+**Outputs:** ES indicator maps, trade-off matrix, summary table by land cover class  
+**Used by workflows:** assess-ecosystem-services  
+
+---
+
+### 12. reproducible-ecology-pipeline
+**Domain:** Provenance tracking, parameter logging, decision audit, reproducibility checklist
+**When to use:** Throughout every project; mandatory before any publication or technical report.
+**Inputs:** Any intermediate or final analytical outputs, code, model parameters
+**Outputs:** Reproducibility checklist, parameter manifest, decision log, audit trail
+**Used by workflows:** all
+
+---
+
+### 13. camera-trap-processing
+**Domain:** Camera trap data processing, detection events, diel activity patterns, temporal overlap
+**When to use:** When processing camera trap image records into structured detection data, computing activity patterns, or generating detection histories for occupancy modeling.
+**Inputs:** Image directory, camera station metadata (GPS, deployment dates), species list, independence threshold (default 30 min)
+**Outputs:** Record table (CSV), detection history matrix, camera operation summary, trap effort table, activity overlap plots
+**Used by workflows:** (standalone; feeds into occupancy-and-detection)
+
+---
+
+### 14. acoustic-monitoring
+**Domain:** Passive acoustic monitoring, soundscape indices (ACI, NDSI, ADI), automated species detection
+**When to use:** When processing audio recordings from AudioMoth, SM4, or similar recorders to compute soundscape indices or detect species via BirdNET or similar classifiers.
+**Inputs:** Audio directory (WAV/FLAC), recording metadata, species list (optional), location coordinates
+**Outputs:** Acoustic indices time series (CSV), detection list with confidence scores, species accumulation curve, soundscape heatmap
+**Used by workflows:** (standalone; pairs with environmental-time-series)
+
+---
+
+### 15. landscape-connectivity
+**Domain:** Habitat connectivity, resistance surfaces, graph-theoretic metrics (IIC, dPC), corridor identification
+**When to use:** To assess landscape connectivity for a focal species, identify wildlife corridors, rank patches by importance, or detect pinchpoints using Circuitscape or graph-based approaches.
+**Inputs:** Habitat patch layer, dispersal distance (m), land cover raster, resistance value table
+**Outputs:** Patch importance metrics (CSV), resistance surface (TIF), connectivity graph plot, pinchpoint map
+**Used by workflows:** (standalone; depends on geoprocessing-for-ecology)
+
+---
+
+### 16. population-viability-analysis
+**Domain:** Matrix population models (Leslie/Lefkovitch), stochastic PVA, extinction risk, IUCN Criterion E
+**When to use:** To project population trajectories, estimate extinction probability, compute lambda/elasticity/sensitivity, or assess a species against IUCN Criterion E quantitative thresholds.
+**Inputs:** Vital rates (survival, fecundity per stage/age), initial population size, time horizon, quasi-extinction threshold
+**Outputs:** Lambda summary, sensitivity/elasticity matrices, stochastic PVA trajectories, extinction probability curve, IUCN Criterion E classification
+**Used by workflows:** (standalone; depends on biostatistics-workbench)
+
+---
+
+### 17. spatial-prioritization
+**Domain:** Systematic conservation planning, prioritizr (ILP), Marxan, Zonation, protected area network design
+**When to use:** To identify priority areas for conservation, design reserve networks with representation targets, calibrate boundary length modifier (BLM), or evaluate existing protected area coverage.
+**Inputs:** Planning unit raster, species/feature distribution rasters, representation targets, cost layer, locked-in/out areas
+**Outputs:** Optimal solution map (TIF), feature representation table, cost summary, irreplaceability map, BLM calibration curve, portfolio frequency map
+**Used by workflows:** (standalone; depends on species-distribution-modeling, ecosystem-services-assessment)
+
+---
+
+## Workflow × Skill Matrix
+
+|  | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 |
+|--|:-:|:-:|:-:|:-:|:-:|:-:|:-:|:-:|:-:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
+| run-sdm-study | ✓ | ✓ | | ✓ | ✓ | ✓ | | | | | | ✓ | | | | | |
+| assess-ecological-impact | ✓ | ✓ | ✓ | | ✓ | | | | ✓ | | | ✓ | | | | | |
+| analyze-community-structure | ✓ | | ✓ | | ✓ | | | ✓ | | | | ✓ | | | | | |
+| build-fire-risk-map | ✓ | ✓ | | ✓ | ✓ | | | | ✓ | ✓ | | | | | | | |
+| run-occupancy-analysis | ✓ | | ✓ | | ✓ | | ✓ | | | | | ✓ | ✓ | | | | |
+| analyze-environmental-change | ✓ | ✓ | | | | | | | ✓ | ✓ | | ✓ | | ✓ | | | |
+| assess-ecosystem-services | ✓ | ✓ | ✓ | | | | | | | | ✓ | ✓ | | | | | ✓ |
+| produce-technical-report | | | | | | | | | | | | ✓ | | | | | |
+| run-multispecies-screening | ✓ | ✓ | | ✓ | ✓ | ✓ | | | | | | | | | | | |
+| run-camera-trap-occupancy | ✓ | | | | ✓ | | ✓ | | | | | ✓ | ✓ | | | | |
+| assess-landscape-connectivity | ✓ | ✓ | | | ✓ | | | | | | | ✓ | | | ✓ | | |
+| run-population-viability | ✓ | | ✓ | | ✓ | | | | | | | ✓ | | | | ✓ | |
+| run-conservation-prioritization | ✓ | ✓ | | | | ✓ | | | | | | ✓ | | | | | ✓ |
+
+---
+
+## New Resources Added in v1.1.0
+
+### skills/species-distribution-modeling/
+| File | Description |
+|---|---|
+| `resources/maxent-calibration-guide.md` | RM × FC grid, OR_AICc criterion, kuenm vs ENMeval comparison |
+| `resources/climate-scenario-preparation.md` | CHELSA/WorldClim sources, SSPs, mandatory preparation pipeline |
+| `scripts/tune_maxnet.R` | ENMeval calibration grid search (35 models), OR_AICc selection, best_maxnet.rds |
+| `scripts/prepare_future_layers.R` | Reproject, crop, compareGeom, layer-name verification for future stacks |
+
+### skills/predictive-modeling-best-practices/
+| File | Description |
+|---|---|
+| `resources/sampling-bias-correction.md` | Target-group background, KDE weighting, environmental filtering, decision table |
+
+### skills/model-validation-and-uncertainty/
+| File | Description |
+|---|---|
+| `resources/extrapolation-risk-guide.md` | MOP, ExDet (NT1/NT2), MESS — comparison table, thresholds, masking protocol |
+| `scripts/extrapolation_risk.R` | MOP + MESS computation, summary CSV, extrapolation plots, severity warnings |
+
+### skills/ecological-data-foundation/
+| File | Description |
+|---|---|
+| `resources/gbif-data-citation-guide.md` | DOI retrieval, occ_search vs occ_download, ODMAP O4 reporting |
+| `scripts/download_from_gbif.R` | Single + batch GBIF download with DOI, rgbif, filters, metadata |
+| `scripts/download_from_gbif.py` | Python equivalent via pygbif |
+
+---
+
+## Advanced Skills (v2.0.0)
+
+### skills/camera-trap-processing/
+| File | Description |
+|---|---|
+| `SKILL.md` | Skill manifest (skill_version: 1.0.0) |
+| `resources/detection-event-definition-guide.md` | Independence threshold by taxon, RAI sensitivity table, decision_log format |
+| `resources/camtrapR-workflow-guide.md` | camtrapR directory structure, recordTable, detectionHistory, cameraOperation |
+| `resources/activity-patterns-reference.md` | Dhat estimators, overlap interpretation, circular statistics, min sample sizes |
+| `scripts/process_camtrap_data.R` | Full camtrapR pipeline: recordTable → detectionHistory → trap effort |
+| `scripts/estimate_activity.R` | Diel activity curves, Dhat4 bootstrap CI, circular stats, overlap plots |
+| `scripts/process_camtrap_data.py` | Pure Python CSV-based summary and detection timeline (no camtrapR dependency) |
+| `examples/example-prompts.md` | 5 prompt scenarios |
+| `examples/leopard_serengeti_example.md` | Full walkthrough: leopard diel activity, temporal overlap, occupancy modelling |
+
+### skills/acoustic-monitoring/
+| File | Description |
+|---|---|
+| `SKILL.md` | Skill manifest (skill_version: 1.0.0) |
+| `resources/acoustic-indices-reference.md` | ACI, BI, NDSI, H, ADI, AEI comparison table, noise sensitivity, soundecology R code |
+| `resources/species-id-tools-comparison.md` | BirdNET vs RavenPro vs others, confidence score table, validation protocol |
+| `resources/soundscape-ecology-guide.md` | NDSI biophony/anthrophony framework, diel control model, rarefaction |
+| `scripts/compute_acoustic_indices.R` | Batch ACI/BI/NDSI/H/ADI/AEI computation, heatmap output, soundecology |
+| `scripts/batch_species_detection.py` | BirdNET batch detection, confidence filtering, species accumulation |
+| `scripts/compute_acoustic_indices.py` | Python ACI/BI/NDSI/H implementation with librosa, matplotlib heatmap |
+| `examples/example-prompts.md` | 5 prompt scenarios |
+| `examples/temperate_forest_birds_example.md` | Full walkthrough: AudioMoth, BirdNET, mixed-effects model, habitat comparison |
+
+### skills/landscape-connectivity/
+| File | Description |
+|---|---|
+| `SKILL.md` | Skill manifest (skill_version: 1.0.0) |
+| `resources/graph-theory-for-ecology.md` | IIC, PC, dIIC, dPC, BC — formulas, R code, interpretation table |
+| `resources/resistance-surface-guide.md` | Resistance value tables, reclassification R code, multi-layer combination |
+| `resources/circuitscape-parameter-guide.md` | INI configuration, R wrapper, pinchpoint extraction, pitfalls |
+| `scripts/connectivity_metrics.R` | IIC, PC, dIIC, dPC, betweenness centrality, graph visualisation |
+| `scripts/resistance_surface.R` | LC reclassification + optional slope + road-proximity resistance layers |
+| `scripts/connectivity_analysis.py` | networkx IIC/PC/dPC, pairwise least-cost paths, matplotlib network plot |
+| `examples/example-prompts.md` | 5 prompt scenarios |
+| `examples/jaguar_mesoamerica_corridor_example.md` | Full walkthrough: MBC corridor analysis, Circuitscape, scenario analysis |
+
+### skills/population-viability-analysis/
+| File | Description |
+|---|---|
+| `SKILL.md` | Skill manifest (skill_version: 1.0.0) |
+| `resources/matrix-model-guide.md` | Leslie vs Lefkovitch, R code for λ/sensitivity/elasticity, uncertainty |
+| `resources/extinction-risk-thresholds.md` | IUCN Criterion E table, 50/500 rule, CV thresholds, stochastic guidance |
+| `resources/sensitivity-elasticity-reference.md` | LTRE, prospective vs retrospective, stochastic elasticity, taxon patterns |
+| `scripts/matrix_pva.R` | Deterministic λ, elasticity heatmap, CV check, projection plot |
+| `scripts/stochastic_pva.R` | Monte Carlo PVA, Beta/Lognormal draws, IUCN Criterion E, MTE |
+| `scripts/pva_analysis.py` | Python λ/sensitivity/elasticity + stochastic PVA, IUCN classification |
+| `examples/example-prompts.md` | 5 prompt scenarios |
+| `examples/african_elephant_pva_example.md` | Full walkthrough: Lefkovitch model, elasticity, drought scenario, poaching threshold |
+
+### skills/spatial-prioritization/
+| File | Description |
+|---|---|
+| `SKILL.md` | Skill manifest (skill_version: 1.0.0) |
+| `resources/prioritizr-formulation-guide.md` | Problem types, BLM calibration, HiGHS solver, irreplaceability, pitfalls |
+| `resources/marxan-vs-prioritizr-comparison.md` | Tool comparison table, when to use each, irreplaceability comparison |
+| `resources/cost-surface-reference.md` | Opportunity cost construction, normalisation, locked-in/out, pitfalls |
+| `resources/representation-targets-guide.md` | 30×30, IUCN-based, log-linear targets, weights, shortfall curves |
+| `scripts/run_prioritization.R` | Min-set or max-coverage ILP, BLM, locked constraints, irreplaceability output |
+| `scripts/prioritization_sensitivity.R` | BLM calibration, target sensitivity, cost scenarios, portfolio frequency |
+| `examples/example-prompts.md` | 5 prompt scenarios |
+| `examples/biodiversity_hotspot_prioritization_example.md` | Full walkthrough: Atlantic Forest 85 species, 30% targets, sensitivity |
+
+### workflows/
+| File | Description |
+|---|---|
+| `run-multispecies-screening/WORKFLOW.md` | 9th workflow — rapid multi-species SDM screening with priority classification |
+| All 8 existing WORKFLOWs | Added `## Decision Points` section with condition → diagnosis → action tables |
+
+---
+
+## Documentation (v2.3.0)
+
+| File | Description |
+|---|---|
+| `docs/theoretical-foundations.md` | Citable justifications for 10 methodological decisions: spatial CV, ensemble, partial ROC, M delimitation, Bray-Curtis, MESH, IIC/PC, OR10, RM calibration, BACI random effects |
+| `docs/global-examples-index.md` | Inventory of all 14 worked examples with coverage analysis (6 continents, 9 biomes, 7 taxonomic groups, 11 analysis types) |
+| `docs/comparison-with-alternatives.md` | Objective comparison vs. Wallace, SDMtoolbox, biomod2, kuenm, ENMTML, Zonation, Vortex |
+
+## New Examples (v2.3.0)
+
+| File | Species / System | Region | Analysis |
+|---|---|---|---|
+| `examples/sdm/wolf_recolonization_europe_example.md` | *Canis lupus* | Western Europe | SDM + recolonization + conflict |
+| `examples/sdm/koala_climate_change_example.md` | *Phascolarctos cinereus* | Eastern Australia | SDM + future SSP + MOP |
+| `examples/community/reef_fish_indopacific_example.md` | Reef fish (987 spp.) | Indo-Pacific | Beta diversity + PERMANOVA |
+| `examples/community/arctic_tundra_vegetation_example.md` | Tundra vegetation | Greenland / Canada | NDVI time series + community shift |
+| `examples/occupancy/snow_leopard_himalayas_example.md` | *Panthera uncia* | Central Himalayas | Occupancy + camera trap |
+| `examples/impact/forest_loss_borneo_timeseries_example.md` | Tropical forest | Borneo | BFAST + MESH + BACI |
+| `examples/reproducible/whittaker_biome_sdm_example.md` | *Vulpes vulpes* | Holarctic | Fully reproducible SDM pipeline |